Method, device and medium for predicting hypotension in hemodialysis

Abstract

The invention relates to the technical field of hemodialysis monitoring, in particular to a method and equipment for predicting hypotension in hemodialysis and a medium thereof. The method comprises the steps of obtaining medical data, and determining a data set after preprocessing; inputting the data set into a statistical model, and based on the definition of dialysis hypotension and a data analysis strategy, obtaining an associated index factor in combination with expert data; inputting the associated index factor into an AI prediction model to obtain a dialysis record result and an AUC value; performing comprehensive evaluation based on the AUC value; if not, adjusting the dialysis hypotension defined value; if the hemodialysis hypotension is greater than the preset value, updating the definition of the hemodialysis hypotension, and optimizing the AI prediction model; if the result is greater than or equal to the expected value, comparing the dialysis record result with the prediction result, and if the condition is met, determining that the current AI prediction model is an optimal prediction model; the dialysis data of the current detection target are input into the optimal prediction model, the probability that the hypotension of the current detection target will occur in the dialysis process is obtained, and the prediction accuracy is high.

Description

Technical Field

[0001] The present invention relates to the technical field of hemodialysis monitoring, and in particular to a method, device and medium for predicting hypotension during hemodialysis. Background Art

[0002] Hypotension is one of the common complications of maintenance hemodialysis (HD) patients during dialysis, which has a significant impact on the patient's survival rate. In severe cases, it can induce arrhythmia, and is therefore one of the main causes of death. Although hemodialysis technology has been developing rapidly, the incidence of hypotension during dialysis has not decreased significantly. With the aging of dialysis patients and the increasing proportion of diabetic nephropathy in the dialysis population, the problem of hypotension during dialysis may become more serious in the future. In order to reduce or improve the incidence of hypotension during dialysis, although there are currently a variety of clinical measures, early intervention based on the prediction of the probability of hypotension occurring during dialysis is still an important research direction and solution.

[0003] The probability prediction of hypotension during dialysis needs to be based on the definition of hypotension during dialysis. However, the existing technology "Prediction Model of Blood Pressure during Hemodialysis Base on DeepLearning" points out that there is no unified consensus on the definition of hypotension during dialysis internationally. This means that different definitions of hypotension during dialysis have a key impact on the establishment of prediction models for the probability of hypotension during dialysis, clinical analysis during dialysis, and clinical evaluation of prediction results (such as accuracy and AUC). For medical staff, different definitions of hypotension during dialysis have different indicators or indicators of change content and process; however, the experience of medical staff in analyzing a large number of patient group indicator data and clinical intervention in discovering hypotension during dialysis and coping is also based on the definition of hypotension during dialysis and related analysis methods, and must also be consistent with the results of related prediction models, which means that the current prediction model of hypotension during dialysis has corresponding limitations.

[0004] For example, the definition of a 30 mmHg drop in systolic blood pressure during dialysis (based on the literature "Analysis of Risk Factors Associated with Hypotension in Maintenance Hemodialysis"), in actual clinical experience, when the pre-dialysis systolic blood pressure is less than 110 mmHg, and the systolic blood pressure drops by more than 22 mmHg during dialysis, the probability of dialysis hypotension is relatively high. In the corresponding clinical cases of dialysis hypotension, when the pre-dialysis systolic blood pressure is less than 120 mmHg, the blood pressure value △SBP dropped during dialysis is almost always less than 30 mmHg. Obviously, the current definition is too strict for the pre-dialysis systolic blood pressure range of less than 120 mmHg. However, for clinical cases of dialysis hypotension in the range of pre-dialysis systolic blood pressure greater than 150 mmHg, the blood pressure value △SBP dropped during dialysis is almost always greater than 30 mmHg. Obviously, the current definition is too loose.

[0005] During dialysis, systolic blood pressure drops by 20 mmHg, accompanied by complications of clinical symptoms. Based on the general definition mentioned in the literature "Prediction Model of Blood Pressure during Hemodialysis Base on DeepLearning", in clinical dialysis records where the pre-dialysis systolic blood pressure is greater than 120 mmHg, there is a probability of about 45% or more that the blood pressure value △SBP during dialysis will drop by more than 20 mmHg. In these record sets that meet the △SBP greater than 20 mmHg, complications with recorded clinical symptoms can be identified as clinical cases of dialysis hypotension. However, in the actual dialysis management process, the records may be incomplete or erroneous, or the authenticity of the clinical complications (whether the complications are caused by dialysis hypotension) may be interfered with by other diseases or medications of the patient. Due to differences in management levels and medical resources of different campuses and hospitals, the current definition has the limitation that there may be differences in the discrimination results during the process of wide promotion and application.

[0006] In summary, different definitions may have different case discriminations for different pre-dialysis systolic blood pressure intervals, which may be too loose, too strict, or have poor controllability. Therefore, the standard represented by the above definition of hypotension cannot well discriminate cases of intradialytic hypotension. This will have a decisive impact on the subsequent intradialytic hypotension prediction model, that is, the intradialytic hypotension prediction model cannot discriminate actual cases of intradialytic hypotension, and the prediction results are inaccurate. Summary of the invention

[0007] The present invention aims to solve at least one of the technical problems existing in the prior art.

[0008] To this end, the present invention provides a method for predicting hypotension during hemodialysis, which can update the definition of dialysis hypotension, predict hypotension during hemodialysis based on the updated definition of dialysis hypotension, and improve the accuracy of prediction.

[0009] A method for predicting hypotension during hemodialysis according to an embodiment of the present invention comprises the following steps:

[0010] S1, obtaining medical data of a target object, and preprocessing the medical data to determine a data set of the target object; wherein the medical data includes HIS data, LIS data, and clinical data of hemodialysis;

[0011] S2, inputting the target object's data set into a statistical model, based on the definition of dialysis hypotension and data analysis strategy, and combining expert data to obtain correlation indicator factors of the definition of hypotension;

[0012] S3, inputting the associated indicator factors of the hypotension definition into the AI ​​prediction model to obtain the dialysis record results and AUC value of the target object;

[0013] S4, performing a comprehensive evaluation based on the AUC value;

[0014] If the AUC value is less than or equal to the preset value, the dialysis hypotension definition value is adjusted and the process returns to step S2;

[0015] If the AUC value is greater than the preset value, the definition of dialysis hypotension is updated, the AI ​​prediction model is optimized, and the process returns to step S2;

[0016] If the AUC value is greater than or equal to the expected value, the dialysis record result of the target object is compared with the predicted result. If the comparison result meets the preset conditions, the current AI prediction model is the optimal prediction model and proceeds to the next step;

[0017] S5, inputting the dialysis data of the current detection target into the optimal prediction model to obtain the probability that the current detection target will experience hypotension during the dialysis process.

[0018] The beneficial effect of the present invention is that a method for predicting hypotension during hemodialysis of the present invention uses an AI prediction model to obtain and update the definition of optimal dialysis hypotension. The AI ​​prediction model predicts hypotension during hemodialysis for patients based on the updated definition of dialysis hypotension, and the obtained results are highly accurate.

[0019] According to one embodiment of the present invention, in step S1, HIS, LIS data and clinical data of hemodialysis are obtained from the hospital information system, laboratory information system and dialysis system respectively;

[0020] Using a time span of days, all the medical data are cleaned and processed to obtain the medical data of the target object every day to form a data set of the target object.

[0021] According to one embodiment of the present invention, step S2 specifically includes the following steps:

[0022] Obtaining the distribution pattern of dialysis hypotension case data in the data set of the target object;

[0023] Combined with expert data, by statistically analyzing the distribution of dialysis hypotension case data in the target object's data set, obtain the characteristic indicator factor variables related to whether dialysis hypotension occurs

[0024] The characteristic indicator factor variables are screened to obtain associated indicator factors for the definition of hypotension.

[0025] According to one embodiment of the present invention, in step S3, the algorithms used by the AI ​​model to process the associated indicator factors defined for hypotension include: logistic regression, linear regression, ridge regression, XGBoost and random forest.

[0026] According to one embodiment of the present invention, updating the definition of dialysis hypotension specifically includes:

[0027] S41, determining the pre-diastolic systolic blood pressure interval based on the target object's data set, and calculating the initial value a of the blood pressure reduction ratio imin And the maximum blood pressure drop ratio a imax , the calculation formula is:

[0028]

[0029] a imin =(ξ imax -90) / ξ imax ;

[0030] where ξ ia is the average value or boundary value of the pre-diastolic systolic pressure interval, ξ imax The maximum value of the dialysis systolic blood pressure interval of the current target object;

[0031] S42, classifying and labeling the data set of the target object as positive and negative examples of hypotension;

[0032] S43, inputting the classified data set of the target object into the AI ​​model to obtain the dialysis record result and AUC value of the target object;

[0033] S44, repeating the step S52 and the step S53 until the blood pressure reduction ratio value α reaches the most stringent standard definition value, obtaining the dialysis record results of the target object whose AUC value is greater than the preset value, and correspondingly collecting the relevant blood pressure reduction ratio value set and calculating the corresponding pressure difference value;

[0034] S45, calculating the difference between the minimum systolic pressure value during dialysis and the systolic pressure value before dialysis in the actual report, finding a maximum value ΔR in the set of pressure reduction difference values ​​Δ, so that all values ​​in the set of pressure reduction values ​​recorded in the actual report of hypotension are greater than ΔR, if the ΔR value is less than the minimum value Δmin in the set of pressure reduction difference values ​​Δ, then ΔR is the minimum value Δmin, and the reduction ratio value corresponding to ΔR is the optimal value α obtained at this time ir , that is, in [α imin ,α imax ] to find an optimal value α ir , obtain the latest definition standard of dialysis hypotension.

[0035] According to one embodiment of the present invention, optimizing the AI ​​prediction model specifically includes the following steps:

[0036] According to the latest definition standard of dialysis hypotension, updating the hypotension label value in the data set of the target object;

[0037] Then, a multi-class machine learning classification algorithm model is used to select the corresponding data of each factor indicator combination in the factor combination set for fitting, and the particle swarm optimization algorithm is used to tune the parameters of each algorithm of the AI ​​model;

[0038] The greedy algorithm is used to filter out the local optimal solutions in multiple dimensions from the fitting results obtained by different algorithms and different hypotension record division strategies;

[0039] Use simulated annealing algorithm to find the local approximate optimal solution of AI model;

[0040] The prediction results of multiple locally optimal algorithm models are used as new features, and the predicted dependent variable, i.e., the qualitative variable (True or False) of whether the dialysis patient has hypotension during the dialysis process, is combined into a new training data set. A new AI model is then constructed by fitting the machine learning and neural network regression model.

[0041] According to one embodiment of the present invention, the data distribution pattern of dialysis hypotension cases in the data set of the target object is obtained:

[0042] Using principal component analysis (PCA) to repeatedly reduce and increase the dimension of the target object's data set;

[0043] A clustering strategy for surface splicing is introduced to extract nonlinear manifold features, and the vectors of the target object data set are quantified using k-means based on unsupervised learning and Euclidean distance measurement;

[0044] The Edited Nearest Neighbours undersampling method algorithm and the Borderline SMO oversampling method algorithm are used to smooth the data set of the target object.

[0045] A computer device according to an embodiment of the present invention includes:

[0046] processor;

[0047] A memory for storing executable instructions;

[0048] Wherein, the processor is used to read the executable instructions from the memory and execute the executable instructions to implement the method for predicting hypotension during hemodialysis as described above.

[0049] According to a computer-readable storage medium of an embodiment of the present invention, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the method for predicting hypotension during hemodialysis as described above.

[0050] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0053] Figure 1 It is a schematic diagram of the method flow of embodiment 1 of the present invention.

[0054] Figure 2 It is a schematic diagram of the overall histogram of the data set data of the first embodiment of the present invention.

[0055] Figure 3 It is a schematic diagram of overall analysis of data distribution according to the first embodiment of the present invention.

[0056] Figure 4 It is a schematic diagram of data density analysis of all indicator factors in the data set of the first embodiment of the present invention.

[0057] Figure 5 It is a schematic diagram of the heat map of the relationship between various characteristic index data of the initial screening in Example 1 of the present invention.

[0058] Figure 6It is a schematic diagram of the definition of hypotension when ΔSBP is 20 mmHg in the first embodiment of the present invention.

[0059] Figure 7 It is a schematic diagram of the definition of hypotension when ΔSBP is 90 mmHg according to the first embodiment of the present invention.

[0060] Figure 8 It is a schematic diagram of low blood pressure definition, in which different pre-dialysis pressure intervals should correspond to an unknown dialysis low blood pressure definition value ΔSBP value according to the first embodiment of the present invention.

[0061] Fig. 9 It is a schematic diagram of the dialysis hypotension pressure difference value according to the first embodiment of the present invention.

[0062] Fig.10 This is a schematic diagram of verifying the definition of dialysis hypotension according to the first embodiment of the present invention.

[0063] Fig.11 It is a schematic diagram of the computer device structure of the second embodiment of the present invention.

[0064] In the figure, 10 is a computer device; 1002 is a processor; 1004 is a memory; and 1006 is a transmission device. DETAILED DESCRIPTION

[0065] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0066] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0067] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0068] Example 1

[0069] The present application embodiment provides a method for predicting hypotension during hemodialysis. Figure 1 As shown, the method comprises the following steps:

[0070] S1, obtaining the medical data of the target object, and preprocessing the medical data to determine the data set of the target object; wherein the medical data includes HIS data, LIS data and clinical data of hemodialysis.

[0071] S2, input the target subject’s data set into the statistical model, based on the definition of dialysis hypotension and data analysis strategy, combined with expert data to obtain the associated indicator factors of the definition of hypotension.

[0072] S3, input the associated indicator factors defined by hypotension into the AI ​​prediction model to obtain the dialysis record results and AUC values ​​of the target subjects.

[0073] S4, conduct a comprehensive assessment based on the AUC value; if the AUC value is less than or equal to the preset value, adjust the definition value of dialysis hypotension △ SBP, return to step S2; if the AUC value is greater than the preset value, the definition of dialysis hypotension is updated, the AI ​​prediction model is optimized, and the process returns to step S2; if the AUC value is greater than or equal to the expected value, the dialysis record result of the target object is compared with the prediction result. If the comparison result meets the preset conditions, the current AI prediction model is the optimal prediction model, and the process proceeds to the next step; in this embodiment, the preset value is set to 75% and the expected value is set to 80%, and the definition of dialysis hypotension can meet the requirements.

[0074] S5, inputting the dialysis data of the current detection target into the optimal prediction model to obtain the probability that the current detection target will experience hypotension during the dialysis process.

[0075] As in step S1, HIS, LIS data and clinical data of hemodialysis can be obtained from the hospital information system, laboratory information system and dialysis system respectively. Among them, HIS data include but are not limited to diagnostic data and medication data; LIS data include but are not limited to laboratory test data (such as blood routine) and other indicator data. Clinical data of hemodialysis include but are not limited to basic dialysis data of patients such as age, weight, systolic blood pressure records before and during dialysis, diastolic blood pressure records before and during dialysis, ultrafiltration, heart rate, diagnosis and complication records.

[0076] Using a time span of days, all medical data are cleaned and processed to obtain the target object's daily medical data and form a data set for the target object. In order to highlight the role of LIS data in supporting the clinical data of hemodialysis, LIS data from the month before dialysis is obtained. In addition, the target object's medical data are the latest historical data retrieved from the system based on the update time; the target object's data set is the medical data corresponding to each patient packaged by patient ID and dialysis time.

[0077] Step S2 specifically includes the following steps:

[0078] Obtain the distribution pattern of dialysis hypotension case data in the target object's data set; specifically including:

[0079] A. Utilize principal component analysis (PCA) to filter and reduce noise on the target object’s dataset by repeated dimension reduction and dimension increase operations.

[0080] B. Introduce clustering to perform surface splicing to extract nonlinear manifold features, use unsupervised learning and k-means based on Euclidean distance measurement to quantize the vectors of the target object's data set, and enhance the characterization ability of the nonlinear data of the target object's data set.

[0081] C. Use the Edited Nearest Neighbours undersampling algorithm and the Borderline SMO oversampling algorithm to smooth the target object’s data set and solve the problem of data skew in the data set.

[0082] Combined with expert opinions, by statistically analyzing the distribution pattern of dialysis hypotension case data in the target object's data set, the characteristic indicator factor variables related to the occurrence of dialysis hypotension are obtained, that is, based on the data distribution of the target factor classification variable (distribution of cases with hypotension), the distribution pattern of dialysis hypotension case data in the target object's data set is analyzed, and the characteristic indicator factor variables related to the occurrence of dialysis hypotension are mined.

[0083] The characteristic indicator factor variables were screened to obtain the associated indicator factors for the definition of hypotension. The specific methods were as follows: SPSS22.0 statistical software and Python platform statsmodels statistical analysis library were used for statistical analysis. The measurement data were expressed as mean ± standard deviation, and the inter-group comparison was performed using t test and variance analysis. The measurement data were expressed as percentages, and the inter-group comparison was performed using χ2 test. The significance test of the characteristic items of categorical data and quantitative targets (ultrafiltration volume) was performed using square T test, and the significance test of the characteristic items of quantitative data and quantitative targets (ultrafiltration volume) was performed using variance test. The optimal characteristic parameter set was determined by multivariate statistical analysis (multiple factors affecting dialysis hypotension at the same time) and multivariate interaction analysis.

[0084] After preliminary research on relevant literature and consultation with relevant experts, expert data was formed. In the early stage, through multi-factor analysis screening based on random forest and XGBoost, age, pre-dialysis weight, dry weight, pre-dialysis systolic blood pressure, pre-dialysis diastolic blood pressure, pre-dialysis pulse, target dehydration, monocytes, eosinophils, pre-dialysis calcium, hematocrit, platelet count, pre-dialysis blood creatinine, pre-dialysis urea nitrogen, fasting blood glucose, and pre-dialysis uric acid were determined as preliminary screening factors related to dialysis hypotension; and the data set corresponding to each factor contained 7098 dialysis record data (dialysis report plus laboratory report data), of which 6353 cases were determined to be non-hypotension records and 745 cases were determined to be hypotension records according to the methods and requirements in 4. The overall histogram of the data can be found in Figure 2 For an overall analysis of data distribution, see Figure 3 , data density analysis of all indicator factors in the data set can be found in Figure 4 , see the heat map of the relationship between the initial screening feature index data Figure 5 .

[0085] In order to verify whether the associated indicator factors of the definition of hypotension are statistically significant, the data of patients in the data set were retrospectively analyzed, and 7098 cases of data were included in the model training set and the model validation set through random balanced sampling. In the model training set, multiple factor indicators affecting dialysis hypotension were first screened out through univariate statistical analysis methods, and the indicator items with statistical significance of P<0.05 were used as variable factors of the model; at the same time, multivariate statistical analysis was performed on the key definition value related to hypotension, that is, the blood pressure drop value (or the lowest blood pressure value available during dialysis), and the indicator factors with clinical significance of P<0.2 were used as model influencing factors. The factor indicators and analysis of the two types of statistics are shown in Table 1:

[0086] Table 1

[0087]

[0088]

[0089] Among them, the test index items with P<0.05 in univariate analysis had statistical significance, and the expert group believed that the test index items with P<0.2 in multivariate analysis had clinical significance. Both univariate and multivariate analysis of eosinophils did not meet the statistical requirements, and they can be considered to be removed from the associated index factor set after the subsequent data indicators are stable.

[0090] In step S3, the algorithm models for AI model processing the associated indicator factors defined for hypotension include: logistic regression, linear regression, ridge regression, XGBoost and random forest. The dialysis record result of the target object is the result of whether the patient will have hypotension during the dialysis process; the ACU value is the area under the ROC curve, which is used to evaluate the performance of the AI ​​model.

[0091] In the existing definition of dialysis hypotension, when the dialysis hypotension definition value △SBP is 20mmHg, the dialysis hypotension definition value △SBP is the minimum value, which can be called the most relaxed definition of hypotension, see Figure 6 shown.

[0092] When the dialysis hypotension definition value △SBP is 90mmHg, the dialysis hypotension definition value △SBP is the minimum value, which can be called the most relaxed definition of hypotension, see Figure 7 shown.

[0093] If the strictest definition standard, i.e., the definition value of dialysis hypotension △SBP is 90mmHg, the vast majority (more than 98%) of dialysis records will be classified as non-low normal pressure. On the contrary, if the loosest definition standard, i.e., the definition value of dialysis hypotension △SBP is 20mmHg, nearly half of the dialysis records will be classified as low normal pressure. Using any value at either end of the dialysis hypotension definition value △SBP value interval as the standard value for defining hypotension does not conform to the medical common sense that the incidence of dialysis hypotension is 10% to 30%. It can be seen that different pre-dialysis pressure intervals should correspond to an unknown dialysis hypotension definition value △SBP value to define hypotension, see Figure 8 shown.

[0094] Therefore, it can be proposed that there is a certain pre-diastolic systolic blood pressure interval ξ a -ξ b A corresponding definition value of dialysis hypotension △SBP can be found in i , so that the interval hypotension recording data conforms to the current definition, which can be understood as the difference between the pre-diastolic systolic pressure and the minimum diastolic pressure in the interval hypotension recording is greater than or equal to the definition-related value △SBP i .

[0095] In addition, in the dialysis records (hypotension complication reports) recorded by medical staff as dialysis hypotension, as the pre-dialysis systolic blood pressure value increases, the corresponding dialysis hypotension definition value △SBP and the pre-dialysis systolic blood pressure value ratio, i.e., the pre-dialysis systolic blood pressure decrease ratio α, will also increase, which is consistent with the above-mentioned reasoning hypothesis, see Fig. 9 shown.

[0096] Therefore, according to the above principle, the relationship between the pre-dialysis systolic pressure value interval and the corresponding pre-dialysis systolic pressure decrease ratio α is shown in the following table:

[0097] Table 2

[0098]

[0099] It should be noted that the analysis interval of pre-dialysis systolic blood pressure less than 90 mmHg is not within the analysis scope of this embodiment.

[0100] The method for assigning the initial value of the blood pressure drop ratio α is based on the most relaxed definition of hypotension (the drop pressure difference is 20 mmHg) based on the average value of the average overall interval of 135 mmHg (90-180 mmHg, the data above 180 mmHg is close to 0 and can be ignored), and the initial drop ratio value corresponding to the average value is 20 mmHg / 135 mmHg. Since the blood pressure drop ratio α increases with the increase of the pre-dialysis systolic blood pressure value, the proportional coefficient can be increased to correct the initial α value, that is, the definition of dialysis hypotension is updated, including:

[0101] S41, determining the pre-diastolic systolic blood pressure interval based on the target object's data set, and calculating the initial value a of the blood pressure reduction ratio imin And the maximum blood pressure drop ratio a imax , the calculation formula is:

[0102]

[0103] a imin =(ξ imax -90) / ξ imax ;

[0104] where ξ ia is the average or boundary value of the pre-diastolic systolic blood pressure interval, ξ imax The maximum value of the dialysis systolic blood pressure interval of the current target object.

[0105] If the AUC value is less than or equal to the preset value, the definition value of dialysis hypotension is adjusted. The specific implementation process is: according to the current definition of dialysis hypotension, the initial value of the blood pressure drop ratio during dialysis is the theoretical minimum value a imin , gradually increase, the step value is 0.01, increase to the theoretical maximum value a imax stop

[0106] S42, classifying and labeling the target object's data set into positive and negative examples of hypotension; the positive and negative examples of hypotension are classified into hypotensive dialysis records and non-hypotensive dialysis records.

[0107] S43, inputting the classified data set of the target object into the AI ​​model to obtain the dialysis record result and AUC value of the target object;

[0108] S44, repeating steps S52 and S53 until the blood pressure reduction ratio value α reaches the most stringent standard definition value, obtaining the dialysis record results of the target object whose AUC value is greater than the preset value, and collecting the relevant blood pressure reduction ratio value set and calculating the corresponding pressure difference value;

[0109] S45, calculating the difference between the minimum systolic blood pressure value during dialysis and the systolic blood pressure value before dialysis in the actual report, finding a maximum value ΔR in the set of blood pressure reduction difference values ​​Δ, so that all values ​​in the blood pressure reduction value set of hypotension records in the actual report are greater than ΔR. If the ΔR value is less than the minimum value Δmin in the set of blood pressure reduction difference values ​​Δ, then ΔR is the minimum value Δmin, and the reduction ratio value corresponding to ΔR is the optimal value α obtained. ir , that is, in [α imin ,α imax ] to find an optimal value α ir , for the latest definition of dialysis hypotension, see Fig.10 shown.

[0110] Optimizing the AI ​​prediction model includes the following steps:

[0111] According to the latest definition standard of dialysis hypotension, update the hypotension label value in the target object's data set;

[0112] Then, a multi-class machine learning classification algorithm model is used to select the corresponding data of each factor indicator combination in the factor combination set for fitting, and the particle swarm optimization algorithm is used to tune the parameters of each algorithm of the AI ​​model;

[0113] The greedy algorithm is used to filter out the local optimal solutions in multiple dimensions from the fitting results obtained by different algorithms and different hypotension record division strategies;

[0114] Use simulated annealing algorithm to find the local approximate optimal solution of AI model;

[0115] The prediction results of multiple locally optimal algorithm models are used as new features, and the predicted dependent variable, i.e., the categorical variable of whether dialysis patients experience hypotension during dialysis, is combined into a new training data set. A new AI model is then constructed by fitting the machine learning and neural network regression model.

[0116] Test verification: 8522 dialysis record data were used as fitting analysis data, of which 15% were test data points and 2359 were recent verification data points. Different blood pressure drop ratio values ​​α were used to generate data record distributions of different types. After the fitting results of the algorithm model, combined with the comprehensive analysis of the corresponding complication report verification and verification, the blood pressure drop ratio values ​​α, accuracy and AUC corresponding to each pre-dialysis systolic blood pressure interval are shown in Table 3:

[0117] Pre-diastolic systolic blood pressure range Optimal blood pressure reduction ratio Validation set accuracy Validation set AUC 90~100mmHg 0.13 96.20％ 0.82 100~110mmHg 0.14 95.29％ 0.81 110~140mmHg 0.20 98.54％ 0.91 140~160mmHg 0.26 96.19％ 0.85 160~180mmHg 0.32 99.11％ 0.95 >180 mmHg 0.34 90.91％ 0.82

[0118] As can be seen from the above table, the updated definition standard of dialysis hypotension meets the definition standard of dialysis hypotension for the current patient group. In the pre-dialysis systolic blood pressure range of 90-180 mmHg where the data are more densely distributed, the AUC value is greater than 0.80, which is in line with the expected value.

[0119] As described in the summary, a method for predicting hypotension during hemodialysis of the present invention obtains and updates the definition of optimal hemodialysis hypotension using an AI prediction model. The AI ​​prediction model predicts hypotension during hemodialysis for patients based on the updated definition of hemodialysis hypotension, and the obtained result is highly accurate. The present invention analyzes and mines out the definition of hemodialysis hypotension applicable to the patient group of the medical unit that meets the current medical management baseline, and models according to the definition, and analyzes the characteristics related to hemodialysis hypotension in the patient group data (including correlation and distribution, etc.), which has good generalizability and easy operability; at the same time, the data set of the present invention is studied based on real clinical data, which is different from the retrospective analysis (or limited to the theoretical research level) after the complete data is interpolated using the data analysis method. Therefore, the analysis and research method of the present invention can be directly applied to various medical units, so as to apply and productize the constructed model, and has high medical application value. 。

[0120] Example 2

[0121] An embodiment of the present application provides a computer device, which includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement a method for predicting hypotension during hemodialysis as provided in the above method embodiment.

[0122] Fig.11 The hardware structure diagram of a device for implementing a method for predicting hypotension in hemodialysis provided in an embodiment of the present application is shown. The device may participate in constituting or include the apparatus or system provided in an embodiment of the present application. Fig.11As shown, the computer device 10 may include one or more processors 1002 (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 1004 for storing data, and a transmission device 1006 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Fig.11 The structure shown is only for illustration and does not limit the structure of the above electronic device. Fig.11 More or fewer components as shown, or with Fig.11 Different configurations shown.

[0123] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer device 10 (or mobile device). As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0124] The memory 1004 can be used to store software programs and modules of application software, such as a program instruction / data storage device corresponding to a method for predicting hypotension in hemodialysis in an embodiment of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 1004, that is, implementing the above-mentioned method. The memory 1004 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 1004 may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computer device 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0125] The transmission device 1006 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer device 10. In one example, the transmission device 1006 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 1006 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0126] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer device 10 (or mobile device).

[0127] Example 3

[0128] An embodiment of the present application also provides a computer-readable storage medium, which can be set in a server to store at least one instruction or at least one program related to a method for predicting hypotension during hemodialysis in a method embodiment. The at least one instruction or the at least one program is loaded and executed by the processor to implement a method for predicting hypotension during hemodialysis provided in the above method embodiment.

[0129] Optionally, in this embodiment, the storage medium may be located in at least one of the multiple network servers of the computer network. Optionally, in this embodiment, the storage medium may include, but is not limited to, various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0130] Example 4

[0131] The embodiment of the present invention further provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes a method for predicting hypotension in hemodialysis provided in the above-mentioned various optional embodiments.

[0132] It should be noted that the above-mentioned sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. The above-mentioned specific embodiments of the present application are described. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. 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, multitasking and parallel processing are also possible or may be advantageous.

[0133] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment and storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0134] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0135] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A method for predicting hypotension during hemodialysis, characterized in that: The method comprises the following steps: S1, obtaining medical data of a target object, and preprocessing the medical data to determine a data set of the target object; wherein the medical data includes HIS data, LIS data, and clinical data of hemodialysis; S2, inputting the target object's data set into a statistical model, based on the definition of dialysis hypotension and data analysis strategy, and combining expert data to obtain correlation indicator factors of the definition of hypotension; S3, inputting the associated indicator factors of the hypotension definition into the AI ​​prediction model to obtain the dialysis record results and AUC value of the target object; S4, performing a comprehensive evaluation based on the AUC value; If the AUC value is less than or equal to the preset value, the dialysis hypotension definition value is adjusted and the process returns to step S2; If the AUC value is greater than the preset value, the definition of dialysis hypotension is updated, the AI ​​prediction model is optimized, and the process returns to step S2; If the AUC value is greater than or equal to the expected value, the dialysis record result of the target object is compared with the predicted result. If the comparison result meets the preset conditions, the current AI prediction model is the optimal prediction model and proceeds to the next step; S5, inputting the dialysis data of the current detection target into the optimal prediction model to obtain the probability that the current detection target will experience hypotension during the dialysis process.

2. The method for predicting hypotension during hemodialysis according to claim 1, wherein: In the step S1, HIS, LIS data and clinical data of hemodialysis are obtained from the hospital information system, laboratory information system and dialysis system respectively; Using a time span of days, all the medical data are cleaned and processed to obtain the medical data of the target object every day to form a data set of the target object.

3. The method for predicting hypotension during hemodialysis according to claim 1, wherein: The step S2 specifically includes the following steps: Obtaining the distribution pattern of dialysis hypotension case data in the data set of the target object; Combined with expert data, by statistically analyzing the distribution of dialysis hypotension case data in the target object's data set, obtain the characteristic indicator factor variables related to whether dialysis hypotension occurs The characteristic indicator factor variables are screened to obtain associated indicator factors for the definition of hypotension.

4. The method for predicting hypotension during hemodialysis according to claim 1, wherein: In step S3, the algorithms used by the AI ​​model to process the associated indicator factors defined for hypotension include: logistic regression, linear regression, ridge regression, XGBoost and random forest.

5. The method for predicting hypotension during hemodialysis according to claim 1, wherein: Updates to the definition of dialysis hypotension include: S41, determining the pre-diastolic systolic blood pressure interval based on the target object's data set, and calculating the initial value a of the blood pressure reduction ratio imin And the maximum blood pressure drop ratio a imax , the calculation formula is: a imin =(ξ imax -90) / x imax ; where ξ ia is the average value or boundary value of the pre-diastolic systolic pressure interval, ξ imax The maximum value of the dialysis systolic blood pressure interval of the current target object; S42, classifying and labeling the data set of the target object as positive and negative examples of hypotension; S43, inputting the classified data set of the target object into the AI ​​model to obtain the dialysis record result and AUC value of the target object; S44, repeating the step S52 and the step S53 until the blood pressure reduction ratio value α reaches the most stringent standard definition value, obtaining the dialysis record results of the target object whose AUC value is greater than the preset value, and correspondingly collecting the relevant blood pressure reduction ratio value set and calculating the corresponding pressure difference value; S45, calculating the difference between the minimum systolic pressure value during dialysis and the systolic pressure value before dialysis in the actual report, finding a maximum value ΔR in the set of pressure reduction difference values ​​Δ, so that all values ​​in the set of pressure reduction values ​​recorded in the actual report of hypotension are greater than ΔR, if the ΔR value is less than the minimum value Δmin in the set of pressure reduction difference values ​​Δ, then ΔR is the minimum value Δmin, and the reduction ratio value corresponding to ΔR is the optimal value α obtained at this time ir， That is, in [α imin ,α imax ] to find an optimal value α ir , obtain the latest definition standard of dialysis hypotension.

6. The method for predicting hypotension during hemodialysis according to claim 5, characterized in that: Optimizing the AI ​​prediction model specifically includes the following steps: According to the latest definition standard of dialysis hypotension, updating the hypotension label value in the data set of the target object; Then, a multi-class machine learning classification algorithm model is used to select the corresponding data of each factor indicator combination in the factor combination set for fitting, and the particle swarm optimization algorithm is used to tune the parameters of each algorithm of the AI ​​model; The greedy algorithm is used to filter out the local optimal solutions in multiple dimensions from the fitting results obtained by different algorithms and different hypotension record division strategies; Use simulated annealing algorithm to find the local approximate optimal solution of AI model; The prediction results of multiple locally optimal algorithm models are used as new features and combined with the predicted dependent variable, i.e., the ultrafiltration volume of MHD patients, to form a new training data set. A new AI model is then constructed through machine learning and neural network regression model fitting.

7. The method for predicting hypotension during hemodialysis according to claim 2, wherein: Obtain the distribution pattern of dialysis hypotension case data in the target object's data set: Using principal component analysis (PCA) to repeatedly reduce and increase the dimension of the target object's data set; A clustering strategy for surface splicing is introduced to extract nonlinear manifold features, and the vectors of the target object data set are quantified using k-means based on unsupervised learning and Euclidean distance measurement; The Edited Nearest Neighbours undersampling method algorithm and the Borderline SMO oversampling method algorithm are used to smooth the data set of the target object.

8. A computer device, characterized in that: include: processor; A memory for storing executable instructions; Wherein, the processor is used to read the executable instructions from the memory, and execute the executable instructions to implement the method for predicting hypotension during hemodialysis as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the method for predicting hypotension during hemodialysis according to any one of claims 1 to 7.

Images