Method and system for predicting acute kidney disease

Abstract

To provide a method and a system for predicting acute renal diseases that accurately predict whether an acute renal disease is in progress.SOLUTION: An acute renal disease prediction system includes an input device, a storage device, a processor, and an output device. A prediction method of the acute renal disease prediction system includes: a step of inputting multiple pieces of immune cell group data, a blood serum creatinine value and a urea nitrogen value, and storing them in the storage device; a step of accessing the storage device, and constructing an acute renal disease prediction model via a decision tree algorithm using the multiple pieces of immune cell group data, the blood serum creatinine value and the urea nitrogen value as parameters; a step of acquiring the immune cell group data, the blood serum creatinine value and the urea nitrogen value that are evaluation objects, and acquiring a determination result procedure of acute renal diseases by executing a determination procedure; and a step of outputting a determination result of acute renal diseases.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a method and a system for predicting acute kidney disease (AKD). In particular, by using a decision tree (DT) algorithm to analyze immune cell group data obtained by a flow cytometer together with serum creatinine (cre) and blood urea nitrogen (BUN) values, it can accurately predict whether a patient will progress to acute kidney disease. The present invention relates to a prediction method and a prediction system capable of achieving this.

Background Art

[0002] Acute kidney disease (AKD) is a severe disease commonly seen in the intensive care unit (ICU). It is not easy to diagnose it at the initial stage of the ICU. The reason is that in clinical judgment, it is necessary for the patient's renal function to be abnormal even 7 days after the onset of acute kidney injury. According to investigations, it has been found that the morbidity rate of acute kidney disease in the intensive care unit accounts for about half of the incidence rate of acute kidney injury. Patients with acute kidney disease may develop into chronic kidney disease. In that case, further medical interventions such as hemodialysis and kidney transplantation are required, so intensive care is necessary.

[0003] Furthermore, during the treatment process, a large number of patients develop sepsis and concurrent acute kidney injury due to bacterial infections through different routes such as wounds, mucous membranes, and the respiratory tract. Also, since the causes of acute kidney disease are complex and change rapidly, how to grasp the treatment timing is a major issue for medical staff. In particular, it is a major issue to grasp whether complications of acute kidney disease will occur and to perform appropriate treatment in advance.

[0004] On the other hand, the clinical indicators previously used to determine whether or not acute kidney injury would develop were serum creatinine and urea nitrogen. In patients with symptoms of acute kidney disease, the concentrations of both are abnormally elevated, so this can be used to determine whether or not a patient will develop acute kidney disease. However, due to the varying physical conditions of different patients, using only serum creatinine and urea nitrogen as clinical indicators has relatively low accuracy.

[0005] Due to the rapid advancement of technology, artificial intelligence is gradually being applied to various medical diagnostic techniques, and the use of computer methods is a powerful tool to support research into related problems in the fields of medicine and biology.

[0006] As described above, in clinical practice, where the presence or absence of acute kidney disease (ACTD) is assessed, there is an urgent need for preventive, effective, and accurate assessment methods so that healthcare professionals can take appropriate preventive measures to reduce the risk of ADD and improve patient survival rates. [Overview of the project] [Problems that the invention aims to solve]

[0007] In view of the conventional problem of evaluating the presence or absence of concomitant acute kidney disease, the object of the present invention is to provide a method and system for predicting acute kidney disease that alleviates the problems of misjudgment due to artificial evaluation and the difficulty in making rapid medical decisions. [Means for solving the problem]

[0008] According to one object of the present invention, a method for predicting acute kidney disease is provided. This prediction method includes the steps of: inputting multiple immune cell group data, serum creatinine levels, and urea nitrogen levels via an input device and storing them in a memory device; accessing the memory device via a processor and constructing an acute kidney disease prediction model using a decision tree algorithm with the multiple immune cell group data, serum creatinine levels, and urea nitrogen levels as parameters; acquiring multiple immune cell group data, serum creatinine levels, and urea nitrogen levels to be evaluated via an input device, performing a determination procedure with a processor to obtain a determination result for acute kidney disease; and accessing the memory device via an output device and outputting the determination result for acute kidney disease. All of these steps are performed by computer software.

[0009] The aforementioned data on multiple immune cell groups, serum creatinine levels, and urea nitrogen levels are all obtained by analyzing peripheral blood samples taken from patients who may have complications of acute kidney disease. The aforementioned data on multiple immune cell groups are obtained by analyzing the peripheral blood samples using a flow cytometer.

[0010] The aforementioned set of multiple immune cell groups consists of 55 specific immune cell group datasets. Therefore, the machine learning model has 55 feature points, and the types of these 55 immune cell group datasets are shown in Table 1 below.

[0011] Table 1 JPEG0007866787000001.jpg249170JPEG0007866787000002.jpg212170

[0012] Then, the aforementioned immune cell population data, serum creatinine levels, and urea nitrogen levels are used as parameters and trained using a decision tree algorithm. Through supervised learning, these parameters are divided into training, validation, and test groups, and the model is continuously trained using a loop approach. When the test group reaches optimal accuracy, the key feature points of the optimal decision tree, namely the key immune cell population data to be combined with serum creatinine and urea nitrogen levels, are identified to obtain an acute kidney disease prediction model.

[0013] Finally, the data of multiple immune cell groups, serum creatinine levels, and urea nitrogen levels to be evaluated are predicted using an acute kidney disease prediction model to obtain a diagnosis of acute kidney disease. Based on the aforementioned diagnosis of acute kidney disease, it is possible to know whether the patient being evaluated has acute kidney disease, allowing healthcare professionals to take appropriate medical action in advance.

[0014] According to another object of the present invention, an acute kidney disease prediction system is provided, comprising an input device, a storage device, a processor, and an output device. The input device receives input from multiple immune cell group data, serum creatinine values ​​and urea nitrogen values, as well as multiple immune cell group data, serum creatinine values ​​and urea nitrogen values ​​to be evaluated. The storage device is connected to the input device and stores the multiple immune cell group data, serum creatinine values ​​and urea nitrogen values, as well as multiple immune cell group data, serum creatinine values ​​and urea nitrogen values. The output device is connected to the storage device and is used to output the results of the acute kidney disease determination. The processor is connected to the storage device and executes a series of instructions to construct an acute kidney disease prediction model using a decision tree algorithm with multiple immune cell group data, serum creatinine values ​​and urea nitrogen values ​​as parameters; executes a determination procedure on the multiple immune cell group data, serum creatinine values ​​and urea nitrogen values ​​to be evaluated based on the acute kidney disease prediction model and obtains the results of the acute kidney disease determination; and accesses the storage device via the output device and outputs the results of the acute kidney disease determination.

[0015] As described above, by using the acute kidney disease prediction method and prediction system of the present invention, it is possible to quickly and accurately determine whether a patient has acute kidney disease, reducing the burden on nurses and doctors, and mitigating problems caused by differences in judgment criteria between nurses and doctors. This allows nurses and doctors to more accurately assess whether a patient has acute kidney disease and to formulate appropriate medical measures more quickly. [Brief explanation of the drawing]

[0016] To further clarify the technical features, content, advantages, and achievable effects of the present invention, the present invention will be described in detail below in the form of embodiments with reference to the accompanying drawings.

[0017] [Figure 1] This figure shows a flowchart of an acute kidney disease prediction method according to one embodiment of the present invention. [Figure 2A] This is a schematic diagram of a decision tree structure obtained using serum creatinine levels and data from multiple immune cell groups as parameters in one embodiment of the present invention. [Figure 2B] This is an analysis diagram of the prediction accuracy of the acute kidney disease prediction model constructed using the decision tree structure shown in Figure 2A. [Figure 3A] This is a schematic diagram of a decision tree structure obtained using urea nitrogen levels and data from multiple immune cell groups as parameters in one embodiment of the present invention. [Figure 3B] This is an analysis diagram of the prediction accuracy of the acute kidney disease prediction model constructed using the decision tree structure shown in Figure 3A. [Figure 4A] This is a schematic diagram of a decision tree structure obtained using serum creatinine and urea nitrogen values ​​as parameters in one embodiment of the present invention. [Figure 4B] This is an analysis diagram of the prediction accuracy of the acute kidney disease prediction model constructed using the decision tree structure shown in Figure 4A. [Figure 5A]In one embodiment of the present invention, it is a schematic diagram of a decision tree structure obtained using serum creatinine value, urea nitrogen value, and immune cells of both Naive Treg and NK CD56d as parameters. [Figure 5B] It is an analysis diagram regarding the prediction accuracy of an acute kidney disease prediction model constructed using the decision tree structure of FIG. 5A. [Figure 6] It is a schematic diagram of an acute kidney disease prediction system according to one embodiment of the present invention.

Embodiments for Carrying Out the Invention

[0018] For the examiner to easily understand the technical features, content, advantages, and achievable effects of the present invention, the present invention will be described in detail in the form of embodiments with reference to the accompanying drawings. The drawings used here are for illustration and to assist in the description, and do not necessarily reflect the actual ratio or exact configuration after the implementation of the present invention. Therefore, the scope of rights of the present invention should not be construed based on the ratio and configuration of the accompanying drawings in actual implementation.

[0019] Unless otherwise defined, all terms (including technical and scientific terms) used in this specification have the same meaning as commonly understood by those with knowledge in the technical field to which the present invention belongs. Furthermore, terms defined as in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related technology and the present invention, and it should be understood that they are not to be interpreted as having an idealized meaning or an overly formal meaning unless clearly defined as such in this specification.

[0020] Refer to FIG. 1 which is a flowchart of an acute kidney disease prediction method according to one embodiment of the present invention. As shown in FIG. 1, the acute kidney disease prediction method includes the following steps (S1 to S4).

[0021] Step S1: Input multiple immune cell group data, serum creatinine levels, and urea nitrogen levels through the input device and save them to the storage device.

[0022] The collected immune cell population data, serum creatinine levels, and urea nitrogen levels are input into the system's storage device via an input device. Here, the input device is not limited to a flow cytometer for acquiring immune cell population data or other analytical instruments capable of acquiring serum creatinine and urea nitrogen levels. Immune cell population data, serum creatinine levels, and urea nitrogen levels stored in the healthcare institution's database can be transmitted via physical lines or files in storage devices. Alternatively, immune cell population data, serum creatinine levels, and urea nitrogen levels can be input into the system's database via wired or wireless network transmission as training data for model building.

[0023] In this study, multiple immune cell group data were obtained by collecting peripheral blood samples from patients at risk of developing acute kidney disease and then analyzing the peripheral blood samples using a flow cytometer. Therefore, all peripheral blood samples from patients at risk of developing acute kidney disease contained multiple immune cell group data, specifically 55 immune cell group data, as described in Table 1 above.

[0024] Step S2: Access the storage device via the processor and construct an acute kidney disease prediction model using a decision tree algorithm with multiple immune cell population data, serum creatinine levels, and urea nitrogen levels as parameters.

[0025] The processor reads multiple immune cell population data, serum creatinine levels, and urea nitrogen levels stored in memory, uses the multiple immune cell population data (the aforementioned 55 immune cell population data) and numerical values ​​such as serum creatinine levels and urea nitrogen levels as training parameters, trains the model using a decision tree algorithm, and continuously trains the model using a loop approach. When the test group achieves optimal accuracy, the optimal key feature points of the decision tree (key immune cell population data) are obtained to obtain an acute kidney disease prediction model.

[0026] Therefore, the parameters obtained from patients potentially at risk of developing acute kidney disease are divided into a training group and a test group based on the total number of patients to construct an acute kidney disease prediction model. The ratio of the training group to the test group is 8:2, i.e., 80% in the training group and 20% in the test group. Since the actual number of patients is 72, the training group will consist of 57 patients and the test group will consist of 15 patients.

[0027] Subsequently, the acute kidney disease (ACTD) prediction model obtained during training was retrained using 106 additional patients who may have a risk of developing ACD as the training group. The prediction accuracy of the retrained ACD prediction model was then verified using 32 additional patients who may have a risk of developing ACD as the validation group. The personal medical information of the 106 patients in the training group and the 32 patients in the validation group is shown in Table 2 below. The 106 patients in the training group were recruited between 2020 and 2021, and the 32 patients in the validation group were recruited in 2022. By increasing the number of people in the training group (sample size), it is possible to further improve the prediction accuracy of the trained ACD prediction model.

[0028] Table 2 JPEG0007866787000003.jpg255170JPEG0007866787000004.jpg25170

[0029] Referring to Figures 2A and 3A, these are schematic diagrams of a decision tree structure obtained using serum creatinine levels and data from multiple immune cell groups as parameters in one embodiment of the present invention, and schematic diagrams of a decision tree structure obtained using urea nitrogen levels and data from multiple immune cell groups as parameters.

[0030] In the schematic diagram of the decision tree structure, Cr represents serum creatinine level, BUN represents blood urea nitrogen level, and gini is the Gini coefficient, which represents the average degree of the classified group (generally, a gini coefficient of 0 represents perfect equality, less than 0.2 represents high equality, 0.2-0.3 represents equality, 0.3-0.4 represents an acceptable range, 0.4-0.6 represents a large disparity, 0.6 or more represents high inequality, and 1.0 represents perfect inequality). `samlpes` is the total number of people, and `value` is the number of people classified by machine learning as potentially having acute kidney disease. `class` is the classification group; in the case of SA-AKD (sepsis-associated acute kidney disease), it is classified as acute kidney disease with sepsis, and if it is not SA-AKD, it is displayed as `Non`.

[0031] For example, in the first level of the decision tree structure in Figure 2A, the classification threshold is a serum creatinine value ≤ 1.185, the total number of people is 57, and the value is [28, 29], where the number 28 on the left represents non-SA-AKD and the number 29 on the right represents SA-AKD, and the Gini coefficient calculated from the ratio of the two is approximately 0.5. Because the number of people with SA-AKD is relatively large, this level is classified as SA-AKD. Next, in the second level, immune cells JPEG0007866787000005.jpg417Treg≦0.002 was the classification threshold, and out of 57 people, 9 were JPEG0007866787000006.jpg417Treg>0.002, value[0,9], gini coefficient0.0, meaning they are completely equal, and all 9 of them are SA-AKD patients, so classification is stopped.The value of the remaining 48 is[28,20], gini coefficient0.486, and there are more non-SA-AKD individuals, so they are classified as Non and classification continues.In this way, classification continues at different thresholds for immune cells until the gini coefficient reaches 0.0, completing the decision tree structure.

[0032] Furthermore, a key feature of the optimal decision tree is that it is extracted using the classification nodes of the decision tree as weights. The higher the classification node, the greater the weight of the immune cells used, and the greater the contribution and importance of whether or not sepsis-related acute kidney disease occurs in the classification.

[0033] From Figures 2A and 3A, the decision tree structure constructed by combining serum creatinine levels and urea nitrogen levels with data from multiple immune cell populations is found in the second level of both. JPEG0007866787000007.jpg contains 417Treg cells, and in the third layer, both contain NK CD56d cells, indicating that these two immune cells play a significant role. Therefore, these two immune cells are used as the main immune cell group data for the optimal decision tree, and serum creatinine and urea nitrogen levels are used as parameters to construct an accurate acute kidney disease prediction model.

[0034] Next, referring to Figures 2B and 3B, these are analysis diagrams of the prediction accuracy of the acute kidney disease prediction model constructed using the decision tree structure shown in Figures 2B and 3B in one embodiment of the present invention. As mentioned above, in these two figures, the accuracy of the acute kidney disease prediction model was confirmed after training with 106 additional patients as the training group and then retraining with 32 additional patients as the validation group. As shown in Figures 2B and 3B, the prediction accuracy of the training group (106 patients) was 89.62%, the sensitivity was 100% and 94.44%, respectively, and the specificity was 84.29% and 87.14%, respectively. However, the prediction accuracy of the validation group (32 patients) was 81.25%, the sensitivity was 100%, and the specificity was 77.78%. From the above, it can be seen that the predictive accuracy of the decision tree structure constructed by combining serum creatinine levels and urea nitrogen levels with data from multiple immune cell groups is all above 80%.

[0035] On the other hand, to confirm that the predictive accuracy of acute kidney disease prediction models constructed by combining serum creatinine or urea nitrogen levels with data from multiple immune cell groups is higher than that of models using only serum creatinine or urea nitrogen levels, which are known clinical indicators of acute kidney disease, we constructed acute kidney disease prediction models using only serum creatinine and urea nitrogen levels as parameters, similarly using a decision tree algorithm, and compared their predictive accuracy.

[0036] Referring to Figures 4A and 4B, these are schematic diagrams of the decision tree structure obtained using serum creatinine and urea nitrogen values ​​as parameters in one embodiment of the present invention, and analytical diagrams of the prediction accuracy of the acute kidney disease prediction model constructed using it.

[0037] As can be seen from the decision tree structure in Figure 4A, it is indeed possible to rapidly classify SA-AKD patients from non-SA-AKD patients using serum creatinine and urea nitrogen levels. However, as can be seen further in Figure 4B, the acute kidney disease prediction model obtained using only serum creatinine and urea nitrogen levels as training parameters had a prediction accuracy of 77.36%, sensitivity of 66.67%, and specificity of 82.86% in the training group (106 patients). In the validation group (32 patients), the prediction accuracy was 75%, sensitivity was 100%, and specificity was 70.37%. From the above, it can be seen that the prediction accuracy of the acute kidney disease prediction model constructed using only serum creatinine and urea nitrogen levels as training parameters is less than 80%, which is lower than the prediction accuracy when combined with data from multiple immune cell groups.

[0038] Furthermore, using a decision tree algorithm, we constructed an acute kidney disease prediction model using multiple immune cell population data, serum creatinine levels, and urea nitrogen levels as parameters, and investigated whether combining these three parameters would further improve the accuracy of the prediction.

[0039] Referring to Figures 5A and 5B, in each embodiment of the present invention, serum creatinine level and urea nitrogen level, further JPEG0007866787000008.jpg417A schematic diagram of the decision tree structure obtained using both Treg and NK CD56d immune cells as parameters, and an analysis diagram of the prediction accuracy of the acute kidney disease prediction model constructed using it.

[0040] As can be seen from Figures 2A and 3A, Because the weight of both Treg and NK CD56d immune cells is relatively large, we will construct an acute kidney disease prediction model by combining these two immune cells with serum creatinine and urea nitrogen levels as training parameters. As can be seen from Figure 5B, serum creatinine and urea nitrogen levels, and further, In an acute kidney disease prediction model constructed using both Treg and NK CD56d immune cells as parameters, the training group (106 patients) showed a prediction accuracy of 88.68%, sensitivity of 94.44%, and specificity of 85.71%. The validation group (32 patients) showed a prediction accuracy of 81.25%, sensitivity of 100%, and specificity of 77.78%. Therefore, by using only two immune cell datasets with larger classification weights and combining them with serum creatinine and urea nitrogen levels, a prediction accuracy of over 80% can be achieved, eliminating the need to simultaneously acquire the aforementioned 55 immune cell datasets as parameters for constructing an acute kidney disease prediction model. This not only improves the speed of model construction but also further enhances the efficiency of peripheral blood sample analysis. All that is needed is to obtain immune cell data for both JPEG0007866787000011.jpg417Treg and NK CD56d.

[0041] Step S3: Data from multiple immune cell groups to be evaluated, serum creatinine levels, and urea nitrogen levels are acquired via the input device, and the processor performs a determination procedure to obtain a result for acute kidney disease.

[0042] Using the acute kidney disease prediction model constructed in step S2, multiple immune cell population data, serum creatinine levels, and urea nitrogen levels are evaluated to obtain a diagnosis of acute kidney disease. The diagnosis indicates whether multiple patients being evaluated will develop acute kidney disease, allowing healthcare professionals to take appropriate medical measures in advance. The input device used here is the same as described above and will not be repeated here.

[0043] Furthermore, to verify the accuracy of acute kidney disease prediction using different algorithms, multiple immune cell population data (i.e., the aforementioned 55 immune cell population data), serum creatinine levels, and urea nitrogen levels were used as parameters, and these parameters were obtained by similarly analyzing peripheral blood samples from 106 patients. For example, acute kidney disease prediction models constructed using other machine learning algorithms such as support vector machines (SVM) and K-nearest neighbors (KNN) were compared with an acute kidney disease prediction model constructed using a decision tree (DT) algorithm, and the prediction accuracy of the three models is shown in Table 3 below.

[0044] Table 3 JPEG0007866787000012.jpg20170

[0045] As can be seen from Table 3, the prediction accuracy of SVM is 70%, and the prediction accuracy of KNN is only 50%, both of which are far lower than the 85% prediction accuracy of the decision tree algorithm. From the above, it can be seen that the acute kidney disease prediction model constructed using the decision tree algorithm is superior to other types of algorithms.

[0046] Step S4: The result of the acute kidney disease diagnosis is output by accessing the storage device via the output device.

[0047] The result of the acute kidney disease diagnosis obtained in step S3 can be output through an output device. The output device disclosed in this embodiment includes various display interfaces such as a computer screen, a display, or a handheld device display.

[0048] Referring to Figure 6, this is a schematic diagram of an acute kidney disease prediction system according to one embodiment of the present invention. As shown in Figure 6, the acute kidney disease prediction system 20 may include an input device 21, a storage device 22, a processor 23, and an output device 24.

[0049] In this embodiment, the input device 21 can consist of a flow cytometer and an analyzer for detecting serum creatinine and urea nitrogen, and analyzes peripheral blood samples taken from patients potentially with acute kidney disease to obtain multiple immune cell group data, serum creatinine values, and urea nitrogen values. In other embodiments, the input device 21 is not limited to a flow cytometer and an analyzer for detecting serum creatinine and urea nitrogen. The input device 21 includes an input interface for an electronic device such as a personal computer, smartphone, or server, including a touchscreen, keyboard, and mouse, and transmits multiple immune cell group data, serum creatinine values, and urea nitrogen values ​​in file form. Alternatively, the historical data may be uploaded and stored in the memory of the storage device 22 via wireless network transmission, wireless communication transmission, or a general wired internet. The memory may include read-only memory, flash memory, disk, or a cloud database.

[0050] Next, the acute kidney disease prediction system 20 accesses the memory device 22 through the processor 23. The processor 23 includes a central processing unit in a computer or server, a graphics processor, a microprocessor, etc., and can also include a multi-core processor or a combination of multiple processors. The processor 23 executes instructions to access multiple immune cell group data, serum creatinine levels, and urea nitrogen levels in the memory device 22, and performs a decision procedure. Specifically, the training procedure uses multiple immune cell group data, serum creatinine levels, and urea nitrogen levels originally stored in the memory device 22 as parameters and performs calculations using a decision tree algorithm to construct a prediction model for acute kidney disease.

[0051] Subsequently, the results for determining acute kidney disease are obtained by calculating the data of multiple immune cell groups, serum creatinine levels, and urea nitrogen levels through a constructed acute kidney disease prediction model according to the determination procedure. The output device 24 accesses the storage device 22 to output the acute kidney disease determination results, and the output device 24 includes various display interfaces, such as a computer screen, display, or handheld device display.

[0052] The acute kidney disease prediction method and prediction system of the present invention significantly reduce the workload of healthcare professionals and physicians, and reduce bias in the diagnosis of whether or not acute kidney disease is present due to human error. Furthermore, it enables rapid and accurate determination of whether or not a patient has acute kidney disease, reducing the burden on nurses and physicians, mitigating the problem of differing judgment criteria among nurses and physicians, and helping nurses and physicians to more accurately assess whether or not acute kidney disease is present and to make appropriate medical decisions more quickly.

[0053] The above are illustrative examples only and are not intended to limit the scope. Equivalent modifications or changes to the present invention should be included in the appended claims without departing from the spirit and scope of the invention. [Explanation of Symbols]

[0054] 20: Acute Kidney Disease Prediction System 21: Input device 22: Storage device 23: Processor 24: Output device S1~S4: Step

Claims

1. A method for predicting acute kidney disease, Step S1 involves inputting multiple immune cell group data, serum creatinine levels, and urea nitrogen levels via an input device and saving them to a storage device. Step S2 involves accessing the storage device via a processor and constructing an acute kidney disease prediction model using a decision tree algorithm with the multiple immune cell group data, serum creatinine value, and urea nitrogen value as parameters. Step S3 involves obtaining data on multiple immune cell groups to be evaluated, serum creatinine levels, and urea nitrogen levels via the input device, and using the acute kidney disease prediction model constructed in step S2, executing a determination procedure on the processor to obtain a determination result for acute kidney disease. Step S4 involves accessing the storage device via the output device and outputting the result of the diagnosis of acute kidney disease. Includes, The aforementioned determination procedure involves confirming whether or not acute kidney disease develops based on the acute kidney disease prediction model. Steps S1 to S4 are all performed by computer software in this method for predicting acute kidney disease.

2. The method for predicting acute kidney disease according to claim 1, wherein the input device comprises a flow cytometer and an analyzer for detecting serum creatinine and urea nitrogen.

3. The aforementioned data on multiple immune cell groups includes B cells, T cells, helper T cells, activated helper T cells (HLADR+), naive helper T cells (CD62L+), naive helper T cells (CD45RA+), memory helper T cells (CD62L-HLADR+), memory helper T cells (CD45RO+), regulatory helper T cells, type 1 helper T cells, naive type 1 helper T cells, memory type 1 helper T cells, and Thai Type 2 helper T cells, naive type 2 helper T cells, memory type 2 helper T cells, regulatory helper T cells, naive regulatory helper T cells, memory type regulatory helper T cells, type 17 helper T cells, naive type 17 helper T cells, memory type 17 helper T cells, type 22 helper T cells, naive type 22 helper T cells, memory type 22 helper T cells, follicular helper T cells, naive follicular helper T cells A method for predicting acute kidney disease according to claim 1, wherein the selection is made from data of 55 immune cell groups, including per T cells, memory-type follicular helper T cells, cytotoxic T cells, activated cytotoxic T cells, naive cytotoxic T cells, memory-type cytotoxic T cells, regulatory cytotoxic T cells, double-positive T cells, double-negative T cells, natural killer cells, natural killer cells CD56b, natural killer cells CD56d, CD56b cells, CD56d cells, natural killer T cells, natural killer T cells CD8, natural killer T cells CD4, double-positive natural killer T cells, double-negative natural killer T cells, CD56b natural killer T cells, CD56d natural killer T cells, CD56bCD16- cells, CD56dCD16- cells, dendritic cells, mature dendritic cells, immature dendritic cells, monocytes, classical monocytes, non-classical monocytes, and intermediate monocytes.

4. The acute kidney disease prediction model includes a decision tree, and the decision tree includes naive regulatory helper T cells ( A method for predicting acute kidney disease according to any one of claims 1 to 3, comprising Treg and natural killer cells CD56d (NK CD56d) as two major immune cell groups.

5. An acute kidney disease prediction system including an input device, a memory device, an output device, and a processor, The input device is used to input multiple immune cell group data, serum creatinine levels and urea nitrogen levels, and multiple immune cell group data, serum creatinine levels and urea nitrogen levels to be evaluated. The storage device is connected to the input device and is used to store the data of the multiple immune cell groups, serum creatinine levels and urea nitrogen levels, and the data of the multiple immune cell groups, serum creatinine levels and urea nitrogen levels to be evaluated. The output device is connected to the storage device and is used to output the results of the diagnosis of acute kidney disease. The processor is connected to the storage device and executes multiple instructions. The steps include constructing an acute kidney disease prediction model using a decision tree algorithm with the aforementioned data of multiple immune cell groups, serum creatinine levels, and urea nitrogen levels as parameters, and The steps include: performing a determination procedure based on the acute kidney disease prediction model on the multiple immune cell group data, serum creatinine levels, and urea nitrogen levels to be evaluated, and obtaining a result for determining acute kidney disease; The process involves accessing the storage device via the output device and outputting the result of the acute kidney disease diagnosis, The aforementioned determination procedure involves confirming whether or not acute kidney disease develops based on the acute kidney disease prediction model. Acute kidney disease prediction system.

6. The acute kidney disease prediction system according to claim 5, wherein the input device comprises a flow cytometer and an analyzer for detecting serum creatinine and urea nitrogen.

Images