A hemodialysis treatment system for a nephrology department

Abstract

The application discloses a hemodialysis processing system for a nephrology department, and belongs to the technical field of hemodialysis, and comprises: a determination module configured to determine at least one target prediction model corresponding to a first patient; a prediction module configured to predict each target dialysis parameter based on each target prediction model when the first patient is subjected to hemodialysis, and obtain a prediction value sequence corresponding to each target dialysis parameter; a trend module configured to determine an actual change trend of each target dialysis parameter, and determine a predicted change trend of each target dialysis parameter based on each prediction value sequence; and a processing module configured to compare each predicted change trend with a corresponding actual change trend, and perform an abnormality reminder for hemodialysis based on a comparison result. The system can accurately monitor abnormal conditions in hemodialysis, so that an abnormality reminder can be timely performed.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of hemodialysis, and particularly relates to a hemodialysis processing system for nephrology. BACKGROUND

[0002] As a key medical means for treating renal failure, hemodialysis removes metabolic waste, excess water and corrects electrolytes in the blood of patients through a hemodialyzer. For nephrology, patients with renal failure, monitoring the hemodialysis process is of great significance to ensure patient safety and dialysis effect when they undergo hemodialysis. The prior art usually sets an alarm threshold for dialysis parameters. If the value of the dialysis parameter monitored during the patient's hemodialysis exceeds the alarm threshold, a warning prompt will occur to remind medical staff to handle it. However, due to the physiological differences between patients, this method of warning prompt when a single value exceeds the alarm threshold has a large error, and it is difficult to accurately monitor whether there is an abnormality in hemodialysis, resulting in difficulty in timely abnormality reminding. Therefore, there is an urgent need for a hemodialysis processing system for nephrology to accurately monitor abnormal conditions in hemodialysis, so as to timely remind the abnormality.

[0003] It should be noted that the information disclosed in the above background section is only used to strengthen the understanding of the background of the present application, and therefore can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0004] The following presents a simplified summary in order to provide a basic understanding of some aspects of the disclosed embodiments. The summary is not an extensive overview of the embodiments described in detail in the following and is not intended to identify key / critical elements or to delineate the scope of the embodiments. Its sole purpose is to present some concepts of the following detailed description in a simplified form.

[0005] The disclosed embodiments provide a hemodialysis processing system for nephrology to accurately monitor abnormal conditions in hemodialysis, so as to timely remind the abnormality.

[0006] In some embodiments, a hemodialysis processing system for nephrology comprises:

[0007] A determination module is configured to determine at least one target prediction model corresponding to a first patient, wherein the target prediction model represents an ARMA model constructed based on a target dialysis parameter and at least one causal dialysis parameter corresponding thereto, and the causal dialysis parameter represents a dialysis parameter having a Granger causality relationship with the target dialysis parameter.

[0008] a prediction module configured to predict each target dialysis parameter based on a corresponding target prediction model when the first patient is undergoing hemodialysis, to obtain a corresponding predicted value sequence of each target dialysis parameter;

[0009] a trend module configured to determine an actual change trend of each target dialysis parameter and a predicted change trend of each target dialysis parameter based on the corresponding predicted value sequence;

[0010] a processing module configured to compare each predicted change trend with the corresponding actual change trend and to provide an abnormality reminder for hemodialysis based on a comparison result.

[0011] The present application has the following beneficial effects:

[0012] The target prediction model corresponding to the first patient is determined by the determining module, which is an ARMA model constructed based on the target dialysis parameter and at least one dialysis parameter having a Granger causality relationship with the target dialysis parameter. The dynamic association and change rule between the target dialysis parameter and each causal dialysis parameter are mined, so that the target dialysis parameter of the first patient can be more accurately predicted by adapting to the individual physiological state of the first patient. Each target dialysis parameter is predicted by using each target prediction model to obtain a corresponding predicted value sequence. The actual change trend and the predicted change trend of each target dialysis parameter are determined by the trend module, so that the processing module can compare each actual change trend with the corresponding predicted change trend.

[0013] The change trend can dynamically reflect the change of the dialysis parameter during the hemodialysis process of the patient, and dynamic monitoring is achieved. In this way, when the first patient is undergoing hemodialysis, the actual change trend of each target dialysis parameter of the first patient is compared with the corresponding predicted change trend, which can more accurately reflect the individual physiological state of the first patient and dynamically monitor the hemodialysis process of the first patient, so that the abnormality in the hemodialysis process can be accurately monitored, and the abnormality reminder can be provided in time.

[0014] The foregoing general description and the following description are only exemplary and explanatory, and are not intended to limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0015] One or more embodiments are exemplarily illustrated by corresponding drawings, which do not constitute a limitation on the embodiments, and elements with the same reference numerals in the drawings represent similar elements, the drawings do not constitute a proportional limitation, and wherein:

[0016] Figure 1 is a structural schematic diagram of a hemodialysis processing system for a nephrology department provided by the present application;

[0017] Figure 2 is another structural schematic diagram of a hemodialysis treatment system for a nephrology department provided by the present application. DETAILED DESCRIPTION

[0018] The characteristics and technical content of the embodiments of the present disclosure are characterized, and the implementation of the embodiments of the present disclosure will be described in detail below in conjunction with the drawings, which are only used for reference and do not limit the embodiments of the present disclosure. In the following technical description, for the convenience of explanation, a plurality of details are provided to provide a full understanding of the disclosed embodiments. However, one or more embodiments can still be implemented without these details. In other cases, the drawings are simplified, and well-known structures and devices can be simplified.

[0019] The terms "first", "second", and the like in the specification and claims of the embodiments of the present disclosure and the above drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present disclosure described herein can be implemented. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion.

[0020] Unless otherwise specified, the term "a plurality of" means two or more.

[0021] In the embodiments of the present disclosure, the character " / " represents an "or" relationship between the objects before and after it. For example, A / B represents: A or B.

[0022] The term "and / or" is a description of the association between objects, which means that there can be three relationships. For example, A and / or B means: A or B, or, A and B, three relationships.

[0023] The term "corresponding" can refer to an association or binding relationship, A corresponding to B means that there is an association or binding relationship between A and B.

[0024] The nephrology department is mainly used for treating patients with kidney disease (such as kidney failure). It removes metabolic waste, excess water, and corrects electrolytes in the blood of the patient through a hemodialysis device to treat the patient with kidney disease.

[0025] In combination Figure 1As shown, the embodiment of the present disclosure provides a hemodialysis processing system for nephrology department, comprising a determination module, a prediction module, a trend module and a processing module. The determination module is configured to determine at least one target prediction model corresponding to a first patient; the target prediction model represents an ARMA model constructed based on a target dialysis parameter and at least one causal dialysis parameter corresponding thereto; the causal dialysis parameter represents a dialysis parameter having a Granger causality relationship with the target dialysis parameter. The prediction module is configured to predict each target dialysis parameter based on each target prediction model when the first patient is undergoing hemodialysis, to obtain a prediction value sequence corresponding to each target dialysis parameter. The trend module is configured to determine an actual change trend of each target dialysis parameter, and determine a predicted change trend of each target dialysis parameter based on each prediction value sequence. The processing module is configured to compare each predicted change trend with the corresponding actual change trend, and perform an abnormality reminder of hemodialysis based on the comparison result.

[0026] By using the hemodialysis processing system for nephrology department provided by the embodiment of the present disclosure, the target prediction model corresponding to the first patient is determined by the determination module, which is an ARMA model constructed based on the target dialysis parameter and at least one dialysis parameter having a Granger causality relationship therewith. By mining the dynamic correlation and change rule between the target dialysis parameter and each causal dialysis parameter, the target dialysis parameter of the first patient can be more accurately predicted to adapt to the individual physiological state of the first patient. Each target dialysis parameter is predicted by using each target prediction model to obtain each corresponding prediction value sequence. The actual change trend and the predicted change trend of each target dialysis parameter are determined by the trend module, so that the processing module compares each actual change trend with the corresponding predicted change trend. The change trend can dynamically reflect the change of the dialysis parameter of the patient during the hemodialysis process, and dynamic monitoring is realized. In this way, when the first patient is undergoing hemodialysis, the actual change trend of each target dialysis parameter of the first patient is compared with the corresponding predicted change trend, which can more accurately monitor the hemodialysis process of the first patient to accurately monitor the abnormal situation in the hemodialysis process, so as to timely perform an abnormality reminder.

[0027] In addition, the work burden of medical staff monitoring the hemodialysis of the patient can be reduced, the real-time performance is stronger, the monitoring result is more accurate, the comprehensive monitoring of each patient is realized, and the safety of the hemodialysis of the patient is improved.

[0028] It can be understood that the first patient in the embodiment of the present disclosure is a kidney disease patient who needs to undergo hemodialysis. The dialysis parameters include blood flow, venous pressure, arterial pressure, dialysate temperature, dialysate flow, ultrafiltration rate, blood pressure, transmembrane pressure, etc., which are mainly dialysis parameters that can be directly obtained through a dialysis machine or an external sensor (such as a blood oxygen detector, a dynamic blood pressure detector) in order to perform high-frequency sampling and subsequent modeling. The predicted value sequence is a set of predicted values arranged in chronological order.

[0029] Preferably, the determining module is specifically configured to determine the at least one target prediction model corresponding to the first patient through the database module if the first patient is an old patient.

[0030] Preferably, the determining module comprises a judging sub-module and a first determining sub-module. The determining module is specifically configured to determine the at least one target prediction model corresponding to the first patient based on the database module through the judging sub-module and the first determining sub-module. The judging sub-module is configured to judge the identity of the first patient. The identity includes an old patient and a new patient. The first determining sub-module is configured to, if the identity of the first patient is an old patient, search for the corresponding at least one target prediction model in a preset first database based on the identity ID of the first patient. The preset first database stores the corresponding relationship between the identity ID and each target prediction model.

[0031] Preferably, the determining module further comprises a second determining sub-module. The second determining sub-module is configured to, if the identity of the first patient is a new patient, determine the corresponding at least one target prediction model based on the medical history information of the first patient. The medical history information includes at least one of dialysis age, age, underlying disease, dialysis frequency, and disease stage.

[0032] When the first patient is an old patient, the first database stores the corresponding at least one target prediction model. At this time, the corresponding at least one target prediction model can be quickly determined by searching in the first database. When the first patient is a new patient, the first database does not store the corresponding target prediction model. At this time, the corresponding at least one target prediction model is determined based on the medical history information of the first patient. The target prediction model can be more consistent with the medical history information of the first patient, so as to accurately predict the dialysis parameters of the first patient.

[0033] It can be understood that the old patient representation in the embodiments of the present disclosure has once received hemodialysis treatment, and the kidney disease patient of at least one target prediction model is stored in the first database. The new patient representation is a kidney disease patient who has not received hemodialysis treatment, or a kidney disease patient who has once received hemodialysis treatment but the target prediction model corresponding to the kidney disease patient is not stored in the first database (for example, the kidney disease patient has once received hemodialysis treatment in A hospital and now receives hemodialysis treatment in B hospital, and the target prediction model corresponding to the kidney disease patient is not stored in the first database of B hospital). The identity ID represents an identity card or other unique identifier.

[0034] In addition, the dialysis age represents the length of time from when the first patient starts to receive hemodialysis treatment to the present, such as "dialysis age 3 years" indicating that the patient has received hemodialysis for 3 years. The underlying disease represents diabetic nephropathy (one of the most common causes), hypertensive renal damage, etc., and the underlying disease directly affects the disease progression, complication type and treatment focus of the hemodialysis patient (for example, diabetic nephropathy patients need to control blood sugar more strictly). The dialysis frequency represents the number of times the kidney disease patient receives hemodialysis per week. The disease stage represents the classification of the current severity of the kidney disease patient, such as stable period, fluctuation period, critical period, etc.

[0035] Preferably, the identity of the first patient is determined, including: searching in a preset second database based on the identity ID of the first patient, if a consistent second patient identity ID is found, then determining the identity of the first patient as an old patient; if a consistent second patient identity ID is not found, then determining the identity of the first patient as a new patient. Wherein, the preset second database stores the corresponding relationship between the identity ID of the first patient and the identity ID of the second patient.

[0036] In this way, by searching in the preset second database based on the identity ID of the first patient, the identity of the first patient can be quickly determined.

[0037] In some embodiments, the first database and the second database can be the same database, so as to optimize the capacity of the database and save data storage resources. The first database and the second database can also be different databases, so as to separate the first database and the second database, thereby protecting the privacy of the patient.

[0038] Preferably, when the identity of the first patient is a new patient, after constructing the target prediction model corresponding to the first patient, the identity ID of the first patient is written into the second database to update the identity ID of the second patient in the second database.

[0039] Preferably, the hemodialysis processing system of the nephrology department further comprises a database module. In combination with Figure 2As shown, the embodiment of the present disclosure provides another blood dialysis processing system for nephrology department, comprising a determining module, a predicting module, a trend module, a processing module and a database module. The database module comprises a first obtaining sub-module, a second obtaining sub-module, a model sub-module and a storage sub-module. The first obtaining sub-module is configured to obtain the identity ID of each second patient. The second obtaining sub-module is configured to obtain at least one target dialysis parameter corresponding to each second patient respectively. The model sub-module is configured to determine at least one causal dialysis parameter corresponding to each target dialysis parameter respectively, and construct a corresponding ARMA model as a target prediction model for each target dialysis parameter based on each target dialysis parameter and each causal dialysis parameter. The storage sub-module is configured to associate the identity ID of each second patient with the corresponding target prediction model, and store in the first database. The second patient represents a nephropathy patient who has received hemodialysis treatment.

[0040] In this way, by obtaining each target dialysis parameter corresponding to each second patient and determining the causal dialysis parameter corresponding to each target dialysis parameter, a corresponding ARMA model is constructed as a target prediction model for each target dialysis parameter based on the Granger causality relationship between each target dialysis parameter and the corresponding at least one causal dialysis parameter, so that the target prediction model can adapt to the individual physiological state of each second patient by mining the dynamic association and change rule between the target dialysis parameter and each causal dialysis parameter, so as to more accurately predict the target dialysis parameter of each second patient. Then, by associating the identity ID of the second patient with the corresponding target prediction model and storing in the first database, the first determining sub-module can quickly find the corresponding at least one target prediction model in the first database.

[0041] It can be understood that for the second patient, the target prediction model associated with it can be updated and stored in the first database based on the change of its disease or dialysis condition.

[0042] Preferably, the second obtaining sub-module comprises a population parameter unit and an individual parameter unit. The population parameter unit is configured to determine at least one target dialysis parameter corresponding to each patient group respectively. The individual parameter unit is configured to determine at least one target dialysis parameter corresponding to each second patient based on at least one target dialysis parameter corresponding to each patient group.

[0043] Preferably, the second obtaining sub-module further comprises a clustering unit. The clustering unit is configured to cluster each second patient based on the medical history information of the plurality of second patients and the K-means clustering algorithm, to obtain a plurality of patient groups.

[0044] Thus, by clustering each second patient so as to group second patients with similar medical history information, at least one target dialysis parameter corresponding to each patient group can be determined in a targeted manner, so as to quickly determine at least one target dialysis parameter corresponding to each second patient.

[0045] It can be understood that the first patient is an old patient, and based on the past hemodialysis treatment, the first patient is also a second patient. Since the number of second patients and the hemodialysis treatment situation can change, each second patient can be clustered periodically to update each patient group.

[0046] Preferably, the clustering unit is specifically configured to: pre-process the medical history information of each second patient to obtain pre-processed medical history information of each second patient; determine a clustering number K (i.e., the number of clusters); randomly select pre-processed medical history information of K second patients as initial clustering centers; calculate distances from pre-processed medical history information of each second patient to each clustering center, and assign each second patient to the nearest clustering center to obtain K clusters; calculate the centroid of each cluster, and update the clustering center to the centroid; repeat the above assignment and updating process until the distance center no longer changes or a preset iteration number is reached, to obtain the final K clusters, i.e., to obtain a plurality of patient groups.

[0047] Preferably, the pre-processing of the medical history information of each second patient to obtain the pre-processed medical history information of each second patient includes: performing standardization processing (Z-score standardization) on continuous variables (such as age and dialysis age) in the medical history information, and converting discrete variables (such as underlying diseases, disease stages, and dialysis frequencies) in the medical history into numerical forms (such as label encoding), so as to pre-process the medical history information of each second patient to obtain the pre-processed medical history information of each second patient.

[0048] Preferably, the determination of the clustering number K includes: calculating total squared errors under different clustering numbers based on an elbow rule, and selecting a clustering number at which the total squared error starts to significantly decrease as the clustering number K.

[0049] Preferably, the calculation of the distance from the pre-processed medical history information of each second patient to each clustering center includes: calculating the Euclidean distance from the pre-processed medical history information of each second patient to each clustering center.

[0050] Preferably, the population parameter unit is configured to obtain, for each patient population, an abnormal frequency of each dialysis parameter being abnormal when the patient population is undergoing hemodialysis, obtain at least one common dialysis parameter, and determine, for each patient population, at least one specific dialysis parameter based on the abnormal frequencies, and determine, for each patient population, at least one target dialysis parameter based on the at least one common dialysis parameter and the at least one specific dialysis parameter.

[0051] For the same patient population, the conditions of each second patient in the patient population are similar. Therefore, by determining the specific dialysis parameters for each patient population, the dialysis parameters that are prone to be abnormal in each patient population can be determined, so as to focus on monitoring. The common dialysis parameters represent the dialysis parameters that are important for all second patients. In this way, by determining the target dialysis parameters for each second patient based on the common dialysis parameters and the specific dialysis parameters of each patient population, the important dialysis parameters that are prone to be abnormal in each patient population and the common important dialysis parameters can be comprehensively considered, so as to achieve the focused monitoring of the dialysis parameters of each second patient, effectively reduce the number of monitored dialysis parameters, and achieve the balance between monitoring effect and monitoring resource saving.

[0052] It can be understood that the common dialysis parameters are determined based on expert experience. The common dialysis parameters include blood pressure, blood flow rate, and ultrafiltration rate.

[0053] Preferably, the population parameter unit is configured to obtain, for each patient population, an abnormal frequency of each dialysis parameter being abnormal when the patient population is undergoing hemodialysis, obtain at least one common dialysis parameter, and determine, for each patient population, at least one specific dialysis parameter based on the abnormal frequencies, and determine, for each patient population, at least one target dialysis parameter based on the at least one common dialysis parameter and the at least one specific dialysis parameter.

[0054] Preferably, the population parameter unit is configured to obtain, for each patient population, an abnormal frequency of each dialysis parameter being abnormal when the patient population is undergoing hemodialysis, obtain at least one common dialysis parameter, and determine, for each patient population, at least one specific dialysis parameter based on the abnormal frequencies, and determine, for each patient population, at least one target dialysis parameter based on the at least one common dialysis parameter and the at least one specific dialysis parameter.

[0055] Preferably, the second determining sub-module is specifically configured to: if the first patient is a new patient, determine a patient group to which the first patient belongs based on the medical history information of the first patient; and determine each target prediction model corresponding to the patient group as each target prediction model corresponding to the first patient, so as to determine at least one target prediction model corresponding to the first patient.

[0056] In this way, when the first patient is a new patient, the target prediction model of the new patient can be determined from the perspective of the patient group by determining the patient group to which the new patient belongs and taking the target prediction model of the patient group as the target prediction model of the new patient.

[0057] In addition, since the target prediction model of the new patient is not a target prediction model constructed based on the hemodialysis condition of the new patient, the accuracy of the target prediction model of the new patient is slightly poorer. Therefore, for the new patient, in addition to the hemodialysis early warning based on the target prediction model, the medical staff will also focus on monitoring, and both methods will be used.

[0058] It can be understood that, in the embodiments of the present disclosure, the target prediction model corresponding to the patient group is determined (only one target prediction model corresponding to the patient group is shown in the embodiments of the present disclosure) by the following method: obtaining the hemodialysis data of each second patient in the patient group and sampling to obtain a sequence set (such as a blood pressure sequence, a blood flow sequence, and an ultrafiltration rate sequence) of each second patient. The mean values of the same dialysis parameters in the sequence sets of all patients are calculated (the mean values of all blood pressure sequences, the mean values of all blood flow sequences, and the mean values of all ultrafiltration rate sequences are calculated), and a mean sequence set is obtained. The mean sequence set includes mean dialysis sequences of multiple dialysis parameters (such as a blood pressure mean sequence, a blood flow mean sequence, and an ultrafiltration rate mean sequence). Each dialysis sequence in the mean sequence set is processed for stationarity, and at least one causal dialysis parameter corresponding to each target dialysis parameter in the patient group is determined based on the Granger causality test method. Based on each target sequence and at least one causal sequence in the patient group, an ARMA model is constructed for each target dialysis parameter in the patient group as the target prediction model corresponding to the patient group; wherein the target sequence corresponding to the patient group represents a mean dialysis sequence in the mean sequence set, and the causal sequence corresponding to the patient group represents a mean dialysis sequence in the mean sequence set corresponding to the causal dialysis parameter corresponding to the patient group.

[0059] Alternatively, preferably, the second determining sub-module is specifically configured to: if the first patient is a new patient, determine a second patient similar to the first patient based on the medical history information of the first patient; and determine each target prediction model corresponding to the second patient as each target prediction model corresponding to the first patient, so as to determine at least one target prediction model corresponding to the first patient.

[0060] In this way, when the first patient is a new patient, a second patient similar to the first patient in terms of medical history information is determined so as to select a target prediction model that is more suitable for the new patient.

[0061] Preferably, the model submodule comprises a sampling unit, a processing unit, a causality unit and a model unit. The sampling unit acquires historical dialysis data of each second patient and performs sampling processing to obtain a dialysis sequence set corresponding to each second patient. The dialysis sequence set comprises dialysis sequences of multiple dialysis parameters. The processing unit is configured to perform stationarity processing on each dialysis sequence in each dialysis sequence set to obtain a stationary sequence set. The stationary sequence set comprises multiple sample sequences, which represent the dialysis sequences after stationarity processing. The causality unit is configured to determine at least one causal dialysis parameter corresponding to each target dialysis parameter based on each stationary sequence set and a Granger causality test method. The model unit is configured to construct an ARMA model for each target dialysis parameter as a target prediction model based on each target sequence and at least one causal sequence. The target sequence represents a sample sequence corresponding to the target dialysis parameter in the stationary sequence set. The causal sequence represents a sample sequence corresponding to the causal dialysis parameter in the stationary sequence set.

[0062] The historical dialysis data is generally data of a complete hemodialysis process, and the historical data is data of regular hemodialysis treatment of the second patient, and the hemodialysis process does not have abnormal conditions. In order to improve the prediction accuracy of the target prediction model, the latest historical dialysis data or the historical dialysis data close to the latest date can be selected; or the historical dialysis data can be updated regularly to update the target prediction model. Stationarity of a sequence is a prerequisite for Granger causality test. Therefore, the stationarity processing is performed on the dialysis sequences in the dialysis sequence set corresponding to each second patient so that the causality unit determines the causal dialysis parameter corresponding to the target dialysis parameter of each second patient based on the Granger causality test method. The model unit constructs an ARMA model based on each target sequence and at least one causal sequence so as to realize accurate prediction of the target dialysis parameter by the lagging effect of the causal dialysis parameter on the target dialysis parameter.

[0063] It can be understood that one second patient corresponds to one dialysis sequence set, one dialysis sequence set corresponds to one stationary sequence set, and one sample sequence in one stationary sequence set corresponds to one dialysis parameter. The sampling points in the dialysis sequence are arranged in time sequence.

[0064] For example, for a second patient, the dialysis parameters include blood flow, venous pressure, arterial pressure, dialysate temperature, dialysate flow, ultrafiltration rate, blood pressure, and transmembrane pressure. The historical dialysis data of the second patient is sampled to obtain a sample point sequence corresponding to each dialysis parameter (which is a sequence arranged in time sequence), i.e., a dialysis sequence (such as a dialysis sequence corresponding to blood flow, a dialysis sequence corresponding to venous pressure, a dialysis sequence corresponding to arterial pressure, a dialysis sequence corresponding to dialysate temperature, a dialysis sequence corresponding to dialysate flow, a dialysis sequence corresponding to ultrafiltration rate, a dialysis sequence corresponding to blood pressure, and a dialysis sequence corresponding to transmembrane pressure). The stability of each dialysis sequence is processed to obtain a sample sequence corresponding to each dialysis sequence. The set of each sample sequence corresponding to all dialysis parameters is the stationary sequence set corresponding to the second patient.

[0065] Preferably, the stability of each dialysis sequence in each dialysis sequence set is processed to obtain each stationary sequence set, including: performing ADF (Augmented Dickey-Fuller test) test on each dialysis sequence to determine whether it is stationary. If yes, it is taken as a sample sequence, and if no, it is processed for stability by difference method (such as first-order difference for a non-stationary dialysis sequence, and then testing whether it is stationary; if not, continue to second-order difference until the sequence is stationary) to obtain a sample sequence.

[0066] Preferably, the causality unit is specifically configured to: take the sample sequence corresponding to the target dialysis parameter in each stationary sequence set as a target sequence, and take the remaining sample sequences other than the target sequence as each basis sequence corresponding to the target sequence. For each target sequence, the following steps are performed: constructing a constrained regression equation for the target sequence, and calculating a first residual sum of squares of the constrained regression equation, and constructing an unconstrained regression equation for the target sequence and each basis sequence corresponding thereto, respectively, and calculating a second residual sum of squares of each unconstrained regression equation, wherein one basis sequence corresponds to one unconstrained regression equation, and calculating the F test statistic of the first residual sum of squares and each second residual sum of squares, respectively, and taking the basis sequence corresponding to the F test statistic satisfying the preset condition as the causal sequence corresponding to the target sequence, and taking the dialysis parameter pointed to by the causal sequence corresponding to the target sequence as the causal dialysis parameter corresponding to the target dialysis parameter.

[0067] In this way, by constructing the constrained regression equation of the target sequence and the unconstrained equation common to the target sequence and one basis sequence, and calculating the F test statistic of the two, if the F test statistic satisfies the preset condition, it means that the basis sequence has a lagging effect on the target sequence and can be used as the causal sequence of the target sequence.

[0068] It can be understood that for the second patient A, its corresponding target dialysis parameters are blood pressure, blood flow and ultrafiltration rate, then the smooth sequence includes blood pressure A=[A1, A2, A3,...An], blood flow B=[B1, B2, B3,...Bn], ultrafiltration rate C=[C1, C2, C3,...Cn], dialysate temperature D=[D1, D2, D3,...Dn], transmembrane pressure E=[E1, E2, E3,...En] and so on. Wherein, then the target dialysis parameter "blood pressure" (the target sequence is [A1, A2, A3,...An]) corresponds to the basic dialysis parameters "blood flow" (the basic sequence [B1, B2, B3,...Bn]), "ultrafiltration rate" (the basic sequence is [C1, C2, C3,...Cn]) and "dialysate temperature" (the basic sequence is [D1, D2, D3,...Dn]) and "transmembrane pressure" (the basic sequence is [E1, E2, E3,...En]). Similarly, the target dialysis parameter "blood flow" (the target sequence [B1, B2, B3,...Bn]), corresponds to the basic dialysis parameters "blood pressure" (the basic sequence [A1, A2, A3,...An]), "ultrafiltration rate" (the basic sequence is [C1, C2, C3,...Cn]) and "dialysate temperature" (the basic sequence is [D1, D2, D3,...Dn]) and "transmembrane pressure" (the basic sequence is [E1, E2, E3,...En]). The target dialysis parameter "ultrafiltration rate" (the target sequence is [C1, C2, C3,...Cn]), corresponds to the basic dialysis parameters "blood pressure" (the basic sequence [A1, A2, A3,...An]), "blood flow" (the basic sequence is [B1, B2, B3,...Bn]), "dialysate temperature" (the basic sequence is [D1, D2, D3,...Dn]) and "transmembrane pressure" (the basic sequence is [E1, E2, E3,...En]).

[0069] Preferably, the target sequence is constructed with a constrained regression equation, including:

[0070]

[0071] Wherein, represents the value of the target sequence at the t time, represents the value of the target sequence lagging i time (also known as lagging i order), represents the intercept term of the target sequence, represents the influence coefficient of the target sequence lagging i time in the constrained regression equation; represents the random error term at the t time in the constrained regression equation; represents the lag order of the target sequence in the constrained regression equation. represents the constrained regression equation.

[0072] It can be understood that in the embodiment, the least square method can be used to estimate parameters, determine the value of , and determine the value of by using the AIC or BIC method, which will not be described herein again.

[0073] The unconstrained regression equation is constructed for the target sequence and each corresponding basic sequence, including:

[0074]

[0075] wherein, represents the value of the target sequence at the t-th moment, represents the value of the target sequence at the i-th lag moment, represents the value of the basic sequence l at the i-th lag moment (each basic sequence corresponds to an unconstrained regression equation); represents the intercept term of the target sequence in the unconstrained regression equation; represents the influence coefficient of the target sequence at the i-th lag moment in the unconstrained regression equation; represents the influence coefficient of the basic sequence l at the i-th lag moment in the unconstrained regression equation, represents the random error term at the t-th moment in the unconstrained regression equation; represents the lag order of the target sequence in the unconstrained regression equation; represents the lag order of the basic sequence in the unconstrained regression equation. represents the unconstrained regression equation.

[0076] It can be understood that in the embodiment, the least square method can be used to estimate parameters, determine the value of and , and determine the value of and by using the AIC or BIC method, which will not be described herein again.

[0077] Preferably, the F test statistic of the first residual sum of squares and each second residual sum of squares is calculated, including:

[0078]

[0079] wherein, F is the F test statistic, is the first residual sum of squares, is the second residual sum of squares, and T is the sample size (i.e., the total number of values in the target sequence), is the total number of parameters in the unconstrained regression equation (i.e., the total number of parameters in the target sequence and each corresponding basic sequence) = + +1).

[0080] Preferably, the basic sequence corresponding to the F test statistic meeting the preset condition is taken as the causal sequence corresponding to the target sequence, comprising: if the F test statistic is greater than a critical value, the null hypothesis (the null hypothesis: assuming that the basic sequence is the causal sequence of the target sequence) is rejected, that is, it is determined that the basic sequence has a significant causal effect on the change of the target sequence, and the basic sequence is the causal sequence corresponding to the target sequence.

[0081] It can be understood that the target dialysis parameter and the causal dialysis parameter corresponding thereto are compared with the same second patient, and the target sequence and all the causal sequences corresponding thereto are compared with the same stationary sequence set.

[0082] Preferably, the model unit is specifically configured to: for each target sequence, perform the following steps: selecting at least one causal sequence from all the causal sequences corresponding to the target sequence to obtain at least one target causal sequence, and constructing a basic ARMA model based on the target sequence and all the target causal sequences corresponding thereto, and determining a target parameter estimation value and a target order of the basic ARMA model based on the target sequence and all the target causal sequences corresponding thereto, and determining an ARMA model as a target prediction model of the corresponding target dialysis parameter based on the target parameter estimation value, the target order and the basic ARMA model.

[0083] In this way, a framework of an ARMA model (Auto-Regressive Moving Average Model) is constructed based on the target sequence and all the causal sequences corresponding thereto, that is, a basic ARMA model. Then, the target sequence and all the causal sequences corresponding thereto are brought into the basic ARMA model to perform parameter value estimation and model order determination, so as to obtain a target prediction model of the target dialysis parameter.

[0084] If the causal sequences corresponding to the target sequence are more, that is, the causal dialysis parameters corresponding to the target dialysis parameter are more, at this time, one or more causal sequences are selected as target causal sequences, and then a basic ARMA model of the target sequence and all the target causal sequences is constructed, which can effectively reduce the calculation amount while ensuring the prediction accuracy. The causal dialysis parameter pointed to by the target causal sequence is the target causal dialysis parameter.

[0085] Preferably, the basic ARMA model is:

[0086]

[0087] wherein, denotes a value of the target sequence at the t th moment, denotes a value of the target sequence at the i th lag moment, denotes a value of the i th lag moment of the target sequence, denotes a value of the i th lag moment of the target causal sequence. The intercept term of the target sequence in the basic ARMA model; The influence coefficient of the target sequence at lag i (i.e., lag order i) in the basic ARMA model; In the basic ARMA model, the influence coefficient of the target causal sequence at lag i is represented. Characterize the random error term at time t in the basic ARMA model; In the basic ARMA model, the lag order of the target sequence is used to characterize the sequence. In the basic ARMA model, the lag order of the target causal sequence is used to characterize the causal sequence. Characterizing the first The identifiers of the target causal sequences, where m represents the total number of target causal sequences. ar represents the underlying ARMA model.

[0088] It is understandable that the lag order of each target causal sequence is not necessarily the same, therefore In fact , ,..., The set. The lag order characterizing the first target causal sequence, Characterizing the lag order of the second target causal sequence, up to The lag order represents the j-th target causal sequence.

[0089] Preferably, the target parameter estimates and target order of the basic ARMA model are determined based on the target sequence and all corresponding target causal sequences. This includes: estimating parameters using nonlinear least squares based on the target sequence and all corresponding target causal sequences to determine the target parameter estimates of the basic ARMA model; and calculating the AIC value corresponding to each order value in a preset order space based on the target sequence and all corresponding target causal sequences, and taking the order value corresponding to the minimum AIC value as the target order of the basic ARMA model.

[0090] In this way, by estimating parameters through nonlinear least squares and by calculating the target parameter estimates and target order obtained by calculating the AIC (Akaike Information Criterion) values ​​corresponding to the preset order space, the ARMA model can be made more stable and its robustness enhanced.

[0091] Understandably, the estimated target parameters include and The estimated value, the order value (i.e., the target order) includes and the value of the order. The order space represents the range of values of the order. In addition, the nonlinear least squares method is used for parameter estimation to obtain the value of and and the Akaike information criterion is used to calculate the value of The specific calculation process of the value is a routine technique, and will not be described here.

[0092] Preferably, the prediction module is specifically configured to determine a corresponding sliding window set based on the target order of each target prediction model; wherein the sliding window set comprises a first sliding window and at least one second sliding window. When the first patient is undergoing hemodialysis, the prediction of each target dialysis parameter is performed by the following method to obtain the prediction value sequence corresponding to each target dialysis parameter: sampling the target dialysis parameter using the corresponding first sliding window to obtain a plurality of first window data, and sampling each target causal dialysis parameter using each second sliding window to obtain a plurality of second window data corresponding to each target causal dialysis parameter; wherein one second sliding window corresponds to one target causal dialysis parameter; aligning each first window data and each second window data to obtain a plurality of aligned sampling window data sets; wherein one sampling window data set comprises one aligned first window data and a plurality of second window data; inputting each sampling window data set into the target prediction model corresponding to the target dialysis parameter to obtain the prediction value corresponding to each sampling window data set output by the target prediction model; and arranging the prediction values in time sequence to obtain the prediction value sequence corresponding to the target dialysis parameter.

[0093] Since the dialysis data of the patient will dynamically change when the patient is undergoing hemodialysis, the dialysis data can be collected and updated in real time by the sliding window method, so as to predict based on the latest dialysis data, thereby improving the prediction accuracy.

[0094] Preferably, the prediction module is specifically configured to determine a corresponding sliding window set based on the target order of each target prediction model; wherein the sliding window set comprises a first sliding window and at least one second sliding window, and comprises: for each target prediction model, the size of the first sliding window is greater than or equal to the lag order of the target sequence in the target prediction model, and the size of each second sliding window is greater than or equal to the lag order of the corresponding target causal sequence in the target prediction model, so as to determine the sliding window set corresponding to the target prediction model.

[0095] For example, the target dialysis parameter is transmembrane pressure, and the corresponding target causal dialysis parameter is blood flow and blood pressure; in the target prediction model ARMA(2, 1, 3) corresponding to the transmembrane pressure, the transmembrane pressure is lagged by 2 orders, the blood flow is lagged by 1 order, and the blood pressure is lagged by 3 orders. The first window data corresponding to the transmembrane pressure includes the measured transmembrane pressures at t-1 and t-2; the second window data corresponding to the blood flow includes the measured blood flow at t-1; and the second window data corresponding to the blood pressure includes the measured blood pressures at t-1, t-2 and t-3. Then, the first window data corresponding to the transmembrane pressure, the second window data corresponding to the blood flow and the second window data corresponding to the blood pressure are taken as a sampling window data set, and the sampling window data set is input into the target prediction model ARMA(2, 1, 3) corresponding to the transmembrane pressure to obtain the predicted transmembrane pressure at t of the target prediction model ARMA(2, 1, 3). The target prediction model ARMA(2, 1, 3) is as follows:

[0096]

[0097] wherein, is the predicted transmembrane pressure at t, is the measured transmembrane pressure at t-1, is the measured transmembrane pressure at t-2; and are the influence coefficients of the transmembrane pressure lagged by 1 order (i.e. lagged by 1 time) and lagged by 2 orders (i.e. lagged by 2 times); is the measured blood flow at t-1; is the influence coefficient of the blood flow lagged by 1 order; , and are the measured blood pressures at t-1, t-2 and t-3; , and are the influence coefficients of the blood pressure lagged by 1 order, lagged by 2 orders and lagged by 3 orders.

[0098] It can be understood that, since the ARMA model constructed by the stationary target sequence and the target causal sequence is used as the target prediction model, the first window data and each second window data collected need to be subjected to the same stationary processing as when the ARMA model is constructed, so as to convert the first window data and each second window data into stationary first window data and each second window data. Then, the stationary first window data and each second window data are input into the target prediction model for prediction. The predicted value is subjected to inverse stationary processing to restore it to the same scale as the sampling data. For example,

[0099] The first window data is stationary (such as first-order difference): for example, the first window data is the transmembrane pressure of the 5th to 7th minute: (Y5 = 120, Y6 = 125, Y7 = 130), the first-order difference is processed to obtain (Z6 = Y6-Y5 = 5), (Z7 = Y7-Y6 = 5); the target prediction model of the transmembrane pressure is used to predict the transmembrane pressure of the 8th minute as Z8 = 5. The 8th minute transmembrane pressure is inverse-differenced (inverse stationary) to restore it to the original scale (i.e. the scale at the time of sampling) Y8 = the measured value Y7 of the last time + the predicted transmembrane pressure of the 8th minute Z8 = 130 + 5 = 135.

[0100] Preferably, the trend module is specifically configured to: obtain a plurality of measured values of each target dialysis parameter to obtain a measured value sequence corresponding to each target dialysis parameter. For each target dialysis parameter: a first linear regression equation is constructed based on the measured value sequence, the slope of the first linear regression equation is taken as the actual change trend of the target dialysis parameter, and a second linear regression equation is constructed based on the predicted value sequence, and the slope of the second linear regression equation is taken as the predicted change trend of the target dialysis parameter.

[0101] In this way, by constructing a linear regression equation, the corresponding slope is obtained, so that the actual change trend and the predicted change trend of each target dialysis parameter can be quickly and accurately determined.

[0102] It can be understood that the first linear regression equation and the second linear regression equation are both linear regression equations. It can be understood that since the patient is undergoing hemodialysis, the hemodialysis data will change dynamically, so the predicted value sequence and the measured value sequence will also gradually increase. Therefore, in order to improve the calculation efficiency, the slopes of the predicted value sequence and the measured value sequence of the latest sliding window can be calculated by the sliding window method, and used as the dynamic update of the actual change trend and the predicted change trend. For example, the prediction window: stores the latest d predicted values (i.e. the predicted values of the last 5 time points) output by the ARMA model. The measured window: stores the latest d measured values (aligned with the prediction window time) of the same period. d can be any value between 5 and 20, or can be adjusted according to actual needs, which is not limited here.

[0103] Preferably, the processing module is specifically configured to: compare each predicted change trend with the corresponding actual change trend. If the absolute value of the difference between any one of the predicted change trends and the corresponding actual change trend in the comparison result is greater than a preset difference threshold, an abnormal hemodialysis reminder is performed.

[0104] In this way, if the absolute value of the difference between any one predicted change trend and the corresponding actual change trend is greater than the preset difference threshold, it indicates that the target dialysis parameter corresponding to the predicted change trend and the actual change trend is abnormal, at which time an abnormality prompt for hemodialysis is performed, which can facilitate medical staff to check and ensure the safety and hemodialysis effect of the first patient during hemodialysis.

[0105] Preferably, the hemodialysis processing system of a nephrology department further comprises a display module. The display module is configured to display each predicted change trend and the corresponding actual change trend of the first patient on a preset display screen.

[0106] In this way, by displaying each predicted change trend and the corresponding actual change trend on a preset display screen (a display screen of a client such as a computer, a mobile phone or a tablet), the monitoring situation of each target dialysis parameter of the first patient can be intuitively displayed.

[0107] Finally, it should be pointed out that the above preferred embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present application.

Claims

1. A hemodialysis treatment system for nephrology, characterized in that, include: The determination module is used to determine at least one target prediction model corresponding to the first patient; wherein, the target prediction model represents an ARMA model constructed based on the target dialysis parameter and the corresponding at least one causal dialysis parameter, and the causal dialysis parameter represents a dialysis parameter that has a Granger causal relationship with the target dialysis parameter; The prediction module is used to predict each target dialysis parameter based on each target prediction model when the first patient undergoes hemodialysis, and obtain the prediction value sequence corresponding to each target dialysis parameter. The trend module is used to determine the actual changing trend of each target dialysis parameter, and to determine the predicted changing trend of each target dialysis parameter based on each predicted value sequence. The processing module is used to compare each predicted trend with the corresponding actual trend and to provide abnormal alerts for hemodialysis based on the comparison results. It also includes a database module; wherein the database module includes: a first acquisition submodule, used to acquire the identity ID of each second patient; a second acquisition submodule, used to acquire at least one target dialysis parameter corresponding to each second patient; a model submodule, used to determine at least one causal dialysis parameter corresponding to each target dialysis parameter, and construct corresponding ARMA models as target prediction models for each target dialysis parameter based on each target dialysis parameter and each causal dialysis parameter; and a storage submodule, used to associate the identity ID of each second patient with the corresponding target prediction models and store them in the first database; The second acquisition submodule includes: a group parameter unit, used to determine at least one target dialysis parameter corresponding to each patient group; and an individual parameter unit, used to determine at least one target dialysis parameter corresponding to each second patient based on at least one target dialysis parameter corresponding to each patient group. The group parameter unit is specifically used to: obtain the abnormal frequency of each dialysis parameter when each patient group is undergoing hemodialysis; obtain at least one general dialysis parameter, and determine at least one specific dialysis parameter for each patient group based on each abnormal frequency; and determine at least one target dialysis parameter for each patient group based on at least one general dialysis parameter and at least one specific dialysis parameter for each patient group. The determination module is specifically used to determine at least one target prediction model corresponding to the first patient if the first patient is an old patient, through the database module.

2. The system according to claim 1, characterized in that, The determining module includes a judgment submodule and a first determining submodule; specifically, the determining module is used to determine at least one target prediction model corresponding to the first patient based on the database module, through the judgment submodule and the first determining submodule, including: The judgment submodule is used to determine the identity of the first patient; the identity includes old patients and new patients. The first determination submodule is used to search in a preset first database based on the first patient's identity ID if the first patient is an old patient, and find at least one corresponding target prediction model; wherein, the preset first database stores the correspondence between identity IDs and each target prediction model.

3. The system according to claim 2, characterized in that, The determining module further includes: The second determination submodule is used to determine at least one target prediction model based on the first patient's medical history information if the first patient is a new patient; wherein the medical history information includes at least one of dialysis age, age, underlying diseases, dialysis frequency and disease stage.

4. The system according to claim 1, characterized in that, The second acquisition submodule further includes: The clustering unit is used to cluster each second patient based on the medical history information of multiple second patients and the K-means clustering algorithm to obtain multiple patient groups.

5. The system according to claim 1, characterized in that, The model submodule includes: The sampling unit acquires historical dialysis data for each second patient and performs sampling processing to obtain a dialysis sequence set corresponding to each second patient; wherein, the dialysis sequence set includes dialysis sequences with multiple dialysis parameters; The processing unit is used to perform stationarity processing on each dialysis sequence within each dialysis sequence set to obtain each stationary sequence set; wherein, the stationary sequence set includes multiple sample sequences, and the sample sequences represent the dialysis sequences that have undergone stationarity processing. The causal unit is used to determine at least one causal dialysis parameter corresponding to each target dialysis parameter based on each set of stationary sequences and the Granger causality test. The model unit is used to construct an ARMA model as a target prediction model for each target dialysis parameter based on each target sequence and at least one corresponding causal sequence; wherein, the target sequence represents the sample sequence corresponding to the target dialysis parameter within the stationary sequence set; and the causal sequence represents the sample sequence corresponding to the causal dialysis parameter within the stationary sequence set.

6. The system according to claim 5, characterized in that, The causal unit is specifically used for: Within each stationary sequence set, the sample sequence corresponding to the target dialysis parameter is taken as the target sequence, and the remaining sample sequences other than the target sequence are taken as the base sequences corresponding to the target sequence. For each target sequence, perform the following steps: Construct a constrained regression equation for the target sequence, and calculate the first sum of squared residuals of the constrained regression equation, and Unconstrained regression equations are constructed for the target sequence and each corresponding base sequence, and the second sum of squared residuals for each unconstrained regression equation is calculated. Each base sequence corresponds to one unconstrained regression equation. Calculate the F-test statistics for the first residual sum of squares and each of the second residual sums of squares, and take the base sequence corresponding to the F-test statistic that meets the preset conditions as the causal sequence corresponding to the target sequence, and take the dialysis parameter pointed to by the causal sequence corresponding to the target sequence as the causal dialysis parameter corresponding to the target dialysis parameter.

7. The system according to claim 6, characterized in that, The model unit is specifically used for: For each target sequence, perform the following steps: Select at least one causal sequence from all causal sequences corresponding to the target sequence to obtain at least one target causal sequence, and Based on the target sequence and all corresponding target causal sequences, a basic ARMA model is constructed, and Based on the target sequence and all corresponding target causal sequences, the estimated target parameters and target order of the basic ARMA model are determined. Based on the target parameter estimates, target order, and basic ARMA model, the ARMA model is determined as the target prediction model for the corresponding target dialysis parameters.

8. The system according to claim 7, characterized in that, The process of determining the target parameter estimates and target order of the basic ARMA model based on the target sequence and all corresponding target causal sequences includes: Based on the target sequence and all corresponding target causal sequences, the nonlinear least squares method is used to estimate the parameters and determine the target parameter estimates of the basic ARMA model. Based on the target sequence and all corresponding target causal sequences, calculate the AIC value corresponding to each order value in the preset order space, and take the order value corresponding to the minimum AIC value as the target order of the basic ARMA model.

9. The system according to any one of claims 1 to 8, characterized in that, The processing module is specifically used for: Compare each predicted trend with the corresponding actual trend; If the absolute value of the difference between any predicted trend and the corresponding actual trend in the comparison results is greater than the preset difference threshold, an abnormality alert for hemodialysis will be issued.

Images