Clinical data analysis and prediction system for critically ill patients based on machine learning

Abstract

The present invention discloses a clinical data analysis and prediction system for critically ill patients based on machine learning. The system first collects clinical data of critically ill patients to be predicted through an information acquisition unit, then predicts and analyzes the clinical data of critically ill patients to be predicted through a server based on a pre-established machine learning model to obtain a first prediction result, and simultaneously obtains a second prediction result of the predictive analysis of the clinical data of critically ill patients to be predicted in a human-participating manner, and then trains the machine learning model based on the first prediction result and the second prediction result, thereby improving the machine learning model, improving the efficiency of data prediction, and ensuring accurate prediction in redundant and complex situations; and then performing hierarchical prediction on the clinical data of critically ill patients to be predicted through at least three levels of pre-trained prediction models, further refining the accuracy of the prediction model and narrowing the range of the degree, which is beneficial for medical staff to make accurate judgments on the criticality of critically ill patients.

Description

Technical Field

[0001] The present invention relates to the field of predictive analysis technology, and in particular to a clinical data analysis and prediction system for critically ill patients based on machine learning. Background Art

[0002] Machine learning is a multi-layer description representation learning, which analyzes and mines existing data to obtain rules and uses these rules to predict unknown data. In recent years, machine learning has been very popular in various fields of society. In the clinical aspects of critically ill patients, machine learning models often predict and analyze the criticality of critically ill patients by establishing a low-latency standardized SAP prediction scoring system. However, since the trained machine learning models have been used in the subsequent prediction and analysis process, there is a lack of improvement in data prediction and analysis. In addition, the scope of clinical data of critically ill patients is complicated, and the degree of quantification of model training is not high, which leads to misjudgment of predictions and the corresponding data prediction is not accurate enough.

[0003] In view of the above technical defects, a solution is now proposed. Summary of the invention

[0004] The purpose of the present invention is to propose a clinical data analysis and prediction system for critically ill patients based on machine learning, which can improve the predictive analysis of the criticality of critically ill patients and solve the problem of low data prediction accuracy.

[0005] In order to achieve the above-mentioned object, the present invention adopts the following technical scheme: a clinical data analysis and prediction system for critically ill patients based on machine learning, comprising a server, an information acquisition unit and an analysis and prediction unit;

[0006] The information acquisition unit sets a sampling period, and acquires the clinical data of the critically ill patients to be predicted in the server according to the preset sampling period; wherein the clinical data of the critically ill patients include the symptoms of the critically ill patients before and after diagnosis and treatment and the disease diagnosis conclusion;

[0007] The disease prediction unit performs predictive analysis on the clinical data of the predicted critically ill patients based on a pre-established machine learning model through a server to obtain a first prediction result, and at the same time obtains a second prediction result of the predictive analysis on the clinical data of the predicted critically ill patients in a manual participation manner; then the machine learning model is trained according to the first prediction result and the second prediction result to obtain an updated prediction model for the next prediction; wherein the first prediction result is a first data set that meets the preset standards and a second data set that does not meet the preset standards, obtained by the server based on the pre-established machine learning model to predict the clinical data of the predicted critically ill patients; the second prediction result is a third data set that meets the preset standards and a fourth data set that does not meet the preset standards, obtained by predicting the clinical data of the predicted critically ill patients in a manual participation manner.

[0008] Furthermore, the symptom prediction unit uses at least three levels of pre-trained prediction models to perform hierarchical prediction on the clinical data of the critically ill patient to be predicted, so as to obtain the prediction results corresponding to the patient's symptom; and also includes a first prediction module, a second prediction module and a third prediction module;

[0009] The first prediction module uses a pre-trained prediction model to perform a first-level prediction on the clinical data of the critically ill patient to be predicted, and obtains the corresponding first-level prediction coefficient;

[0010] The second prediction module uses a pre-trained prediction model to perform a second-level prediction on the clinical data of the critically ill patients to be predicted based on the first-level prediction coefficient, and obtains the corresponding second-level prediction coefficient;

[0011] The third prediction module uses a pre-trained prediction model to perform a third-level prediction on the clinical data of the critically ill patients to be predicted based on the second prediction coefficient, and obtains the corresponding third-level prediction coefficient.

[0012] Furthermore, the specific process of training the machine learning model according to the first prediction result and the second prediction result is:

[0013] First, perform data misjudgment analysis based on the first prediction result and the second prediction result, assign corresponding hash values ​​based on the misjudgment analysis results, and match unique data identifiers for corresponding data, specifically: misjudge data that meets the preset standard based on the pre-established machine learning model as data that does not meet the preset standard, and generate a first misjudgment identifier; misjudge data that does not meet the preset standard based on the pre-established machine learning model as data that meets the preset standard, and generate a second misjudgment identifier; misjudge data that cannot be identified based on the pre-established machine learning model but actually meets the preset standard, and generate a third misjudgment identifier; misjudge data that cannot be identified based on the pre-established machine learning model but actually does not meet the preset standard, and generate a fourth misjudgment identifier; wherein, data that cannot be identified based on the pre-established machine learning model but actually does not meet the preset standard is marked as a fifth data set;

[0014] Acquire data that does not exist in the first data set in the third data set to obtain a sixth data set; and simultaneously acquire data that does not exist in the third data set in the first data set to obtain a seventh data set; and then use the sixth data set and the first identifier, and the seventh data set and the second identifier as input information to train a machine learning model to obtain an updated first prediction model;

[0015] Acquire data that does not exist in the fourth data set in the second data set to obtain an eighth data set; and simultaneously acquire data that does not exist in the second data set in the fourth data set to obtain a ninth data set; and then use the eighth data set and the first identifier, and the ninth data set and the second identifier as data information to train a machine learning model to obtain an updated second prediction model;

[0016] Acquire data that does not exist in the first data set and the fifth data set in the third data set to obtain a tenth data set; acquire data that does not exist in the third data set and the fifth data set in the first data set to obtain an eleventh data set; acquire data that exists in the third data set in the fifth data set to obtain a twelfth data set; acquire data that exists in the fourth data set in the fifth data set to obtain a thirteenth data set; and then use the tenth data set and the first identifier, the eleventh data set and the second identifier, the twelfth data set and the third identifier, and the thirteenth data set and the fourth identifier as input information to train a machine learning model to obtain an updated third prediction model;

[0017] Acquire data from the second data set that does not exist in the fourth data set and the fifth data set to obtain a fourteenth data set; acquire data from the fourth data set that does not exist in the second data set and the fifth data set to obtain a fifteenth data set; acquire the twelfth data set and the thirteenth data set at the same time, and then use the fourteenth data set and the first identifier, the fifteenth data set and the second identifier, the twelfth data set and the third identifier, and the thirteenth data set and the fourth identifier as input information to train the machine learning model to obtain an updated fourth prediction model.

[0018] Furthermore, the pre-trained first prediction model is used to perform the first-level prediction on the clinical data of the critically ill patients to be predicted. The specific working process is as follows:

[0019] A processing cycle T1 is set, and the clinical data of critically ill patients to be predicted in the processing cycle T1 is marked as the first target data, and the first target data is input into the first prediction model to obtain a first prediction result, and the first prediction result is converted and analyzed to generate a first prediction coefficient; wherein the first prediction result describes the degree of the diagnosis result for the critically ill patients, and the first prediction coefficient is the degree value obtained according to the degree of the disease corresponding to the first prediction result.

[0020] Furthermore, the pre-trained second prediction model is used to perform the second-level prediction on the clinical data of the critically ill patients to be predicted. The specific working process is as follows:

[0021] Based on the first prediction result, a processing cycle T2 is set, and the clinical data of the critically ill patients to be predicted in the processing cycle T2 is marked as the second target data, the second target data is input into the second prediction model, a second prediction result is obtained, and the second prediction result is converted and analyzed to generate a second prediction coefficient; wherein the second prediction result is the result of subdividing the degree of the disease based on the first prediction result, the second prediction result describes the degree of the diagnostic results for the critically ill patients, and the second prediction coefficient is the degree value obtained according to the degree of the disease corresponding to the second prediction result.

[0022] Furthermore, the third level prediction model that has been pre-trained is used to perform the third level prediction on the clinical data of the critically ill patients to be predicted. The specific working process is as follows:

[0023] Based on the second prediction result, a processing cycle T3 is set, and the clinical data of the critically ill patients to be predicted in the processing cycle T3 is marked as third target data, the third target data is input into the third prediction model, a third prediction result is obtained, and the third prediction result is converted and analyzed to generate a third prediction coefficient; wherein the third prediction result is the result of a subdivision of the degree of the disease based on the second prediction result, the third prediction result describes the degree of the diagnostic results for the critically ill patients, and the third prediction coefficient is the degree value obtained according to the degree of the disease corresponding to the third prediction result.

[0024] Furthermore, based on the first-level prediction coefficient, the second-level prediction coefficient and the third-level prediction coefficient, a pre-trained fourth prediction model is used to conduct a comprehensive prediction analysis on the clinical data of the critically ill patients to be predicted. The specific working process is as follows:

[0025] First, the first prediction coefficient, the second prediction coefficient and the third prediction coefficient are marked as yc1, yc2 and yc3 respectively, and then the processing period T4 is set, and a four-dimensional comprehensive prediction vector <first prediction coefficient yc1, second prediction coefficient yc2, third prediction coefficient yc3, processing period T4> is constructed, and then the first prediction coefficient yc1 in the processing period T4 is obtained, and then the average value and fluctuation value corresponding to the first prediction coefficient yc1, the second prediction coefficient yc2 and the third prediction coefficient yc3 in the processing period T4 are statistically analyzed. For any time period Ta in the processing period T4, the average values ​​of the first prediction coefficient yc1, the second prediction coefficient yc2 and the third prediction coefficient yc3 marked in the time period Ta are Pj respectively. yc1 , Pj yc2 and Pj yc3 , and the fluctuation values ​​are Bd yc1 , Bd yc2 and Bd yc3 , and construct judgment vectors according to different combinations;

[0026] The judgment vector is then substituted into the fourth prediction model. When the result corresponding to any judgment vector is smaller than the threshold value stored internally in the server, an emergency signal is output to determine that the critically ill patient is at a critical level.

[0027] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0028] The present invention first collects clinical data of critically ill patients to be predicted through an information acquisition unit, and then performs predictive analysis on the clinical data of critically ill patients to be predicted through a server based on a pre-established machine learning model to obtain a first prediction result, and at the same time obtains a second prediction result of predictive analysis on the clinical data of critically ill patients to be predicted in a human-participating manner, and then trains the machine learning model based on the first prediction result and the second prediction result, thereby achieving an improvement in the machine learning model, improving the efficiency of data prediction to a certain extent, and ensuring accurate prediction in redundant and complex situations; and then performing hierarchical prediction on the clinical data of critically ill patients to be predicted through at least three levels of pre-trained prediction models, further refining the accuracy of the prediction model and narrowing the range of the degree, which is conducive to medical staff making accurate judgments on the criticality of critically ill patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 The structural schematic diagram of the present invention is shown;

[0030] Figure 2 A schematic diagram of the process of the present invention is shown. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] Embodiment 1:

[0033] like Figure 1-Figure 2 As shown, a clinical data analysis and prediction system for critically ill patients based on machine learning includes a server, an information acquisition unit, and an analysis and prediction unit;

[0034] The specific working steps are as follows:

[0035] S1, the information acquisition unit sets a sampling period, and acquires the clinical data of the critically ill patient to be predicted in the server according to the preset sampling period; wherein the clinical data of the critically ill patient includes the symptoms of the critically ill patient before and after diagnosis and treatment and the disease diagnosis conclusion; it should be noted that the symptoms of the critically ill patient after diagnosis and treatment include but are not limited to the diagnosis and treatment CT images and clinical manifestations of the critically ill patient, and the disease diagnosis conclusion includes but is not limited to the diagnosis and treatment parameters of the disease;

[0036] S2, the symptom prediction unit performs predictive analysis on the clinical data of the critically ill patient to be predicted based on the pre-established machine learning model through the server to obtain a first prediction result, and at the same time obtains a second prediction result of the predictive analysis on the clinical data of the critically ill patient to be predicted in a human-involved manner; and then trains the machine learning model according to the first prediction result and the second prediction result to obtain an updated prediction model for the next prediction; and the unit uses at least three levels of pre-trained prediction models to perform hierarchical prediction on the clinical data of the critically ill patient to be predicted to obtain the prediction result corresponding to the patient's symptom; and also includes a first prediction module, a second prediction module and a third prediction module;

[0037] Among them, the first prediction result is a first data set that meets the preset standard and a second data set that does not meet the preset standard, which are obtained by the server based on the pre-established machine learning model to predict the clinical data of the critically ill patients to be predicted; the second prediction result is a third data set that meets the preset standard and a fourth data set that does not meet the preset standard, which are obtained by predicting the clinical data of the critically ill patients to be predicted in a manual participation manner;

[0038] The specific process of training the machine learning model according to the first prediction result and the second prediction result is:

[0039] Sa, first perform data misjudgment analysis according to the first prediction result and the second prediction result, assign corresponding hash values ​​according to the misjudgment analysis results, and match unique data identifiers for corresponding data, specifically: misjudge data that meets the preset standard based on the pre-established machine learning model as data that does not meet the preset standard, and generate a first misjudgment identifier; misjudge data that does not meet the preset standard based on the pre-established machine learning model as data that meets the preset standard, and generate a second misjudgment identifier; misjudge data that cannot be distinguished based on the pre-established machine learning model but actually meets the preset standard, and generate a third misjudgment identifier; misjudge data that cannot be distinguished based on the pre-established machine learning model but actually does not meet the preset standard, and generate a fourth misjudgment identifier; wherein, data that cannot be judged based on the pre-established machine learning model but actually meets the preset standard is marked as a fifth data set;

[0040] Sb, obtaining data that does not exist in the first data set in the third data set to obtain a sixth data set; and obtaining data that does not exist in the third data set in the first data set to obtain a seventh data set; and then using the sixth data set and the first identifier, and the seventh data set and the second identifier as input information to train a machine learning model to obtain an updated first prediction model;

[0041] Sc, obtaining data that does not exist in the fourth data set from the second data set to obtain an eighth data set; and obtaining data that does not exist in the second data set from the fourth data set to obtain a ninth data set; and then using the eighth data set and the first identifier, and the ninth data set and the second identifier as data information to train a machine learning model to obtain an updated second prediction model;

[0042] Sd, obtain data that does not exist in the first data set and the fifth data set in the third data set to obtain a tenth data set; simultaneously obtain data that does not exist in the third data set and the fifth data set in the first data set to obtain an eleventh data set; simultaneously obtain data that exists in the third data set in the fifth data set to obtain a twelfth data set; simultaneously obtain data that exists in the fourth data set in the fifth data set to obtain a thirteenth data set; and then use the tenth data set and the first identifier, the eleventh data set and the second identifier, the twelfth data set and the third identifier, and the thirteenth data set and the fourth identifier as input information to train a machine learning model to obtain an updated third prediction model;

[0043] Se, obtain data that does not exist in the fourth data set and the fifth data set in the second data set to obtain a fourteenth data set; at the same time, obtain data that does not exist in the second data set and the fifth data set in the fourth data set to obtain a fifteenth data set; at the same time, the twelfth data set and the thirteenth data set, and then use the fourteenth data set and the first identifier, the fifteenth data set and the second identifier, the twelfth data set and the third identifier, and the thirteenth data set and the fourth identifier as input information to train the machine learning model to obtain an updated fourth prediction model;

[0044] Among them, the sixth data set, the eighth data set, the tenth data set and the fourteenth data set are all based on the pre-established machine learning model to misjudge the data that meets the preset standard as data that does not meet the preset standard; the seventh data set, the ninth data set, the eleventh data set and the fifteenth data set are all based on the pre-established machine learning model to misjudge the data that does not meet the preset standard as data that meets the preset standard;

[0045] S2-1, the first prediction module uses a pre-trained prediction model to perform a first-level prediction on the clinical data of the critically ill patient to be predicted, and obtains the corresponding first-level prediction coefficient;

[0046] The specific working process is:

[0047] A processing cycle T1 is set, clinical data of critically ill patients to be predicted in the processing cycle T1 is marked as first target data, the first target data is input into a first prediction model, a first prediction result is obtained, and the first prediction result is converted and analyzed to generate a first prediction coefficient; wherein the first prediction result describes the degree of the diagnosis result of the critically ill patient, and the first prediction coefficient is a degree value obtained according to the degree of the disease corresponding to the first prediction result;

[0048] S2-2, the second prediction module uses a pre-trained prediction model to perform a second-level prediction on the clinical data of the critically ill patient to be predicted based on the first-level prediction coefficient, and obtains the corresponding second-level prediction coefficient;

[0049] The specific working process is:

[0050] Based on the first prediction result, a processing cycle T2 is set, and the clinical data of the critically ill patient to be predicted in the processing cycle T2 is marked as the second target data, the second target data is input into the second prediction model, a second prediction result is obtained, and the second prediction result is converted and analyzed to generate a second prediction coefficient; wherein the second prediction result is a result of subdividing the degree of the disease based on the first prediction result, the second prediction result describes the degree of the diagnosis result of the critically ill patient, and the second prediction coefficient is a degree value obtained according to the degree of the disease corresponding to the second prediction result;

[0051] S2-3, the third prediction module uses a pre-trained prediction model to perform a third-level prediction on the clinical data of the critically ill patient to be predicted based on the second prediction coefficient, and obtains a corresponding third-level prediction coefficient;

[0052] The specific working process is:

[0053] Based on the second prediction result, a processing cycle T3 is set, and the clinical data of the critically ill patient to be predicted in the processing cycle T3 is marked as the third target data, the third target data is input into the third prediction model, a third prediction result is obtained, and the third prediction result is converted and analyzed to generate a third prediction coefficient; wherein the third prediction result is the result of subdividing the degree of the disease based on the second prediction result, the third prediction result describes the degree of the diagnosis result of the critically ill patient, and the third prediction coefficient is the degree value obtained according to the degree of the disease corresponding to the third prediction result;

[0054] It should be noted that, in this embodiment, the degree of the disease severity is analyzed by first setting the level of the degree value corresponding to the first prediction result. When the degree value = 0, it indicates that there are no severe symptoms; when the degree value = 1, it indicates that the severe symptoms are mild; when the degree value = 2, it indicates that the severe symptoms are moderate; when the degree value = 3, it indicates that the severe symptoms are severe; based on the first prediction result, when the second prediction result is determined to be a severe symptom in the degree value [2,3], a detailed degree analysis is performed based on the severity of the severe symptoms, such as: when the degree value = 2.0, it indicates that it is in the severe latent early stage of severe symptoms; when the degree value = 2.50, it indicates that it is in the severe latent middle stage of severe symptoms; when the degree value = 2.50, it indicates that it is in the severe latent latent stage of severe symptoms; based on the second prediction result, when the third prediction result is determined to be a severe latent latent state of severe symptoms, the larger the degree value, the greater the corresponding degree of urgency;

[0055] S2-4, based on the first-level prediction coefficient, the second-level prediction coefficient and the third-level prediction coefficient, a pre-trained fourth prediction model is used to conduct a comprehensive prediction analysis on the clinical data of the critically ill patients to obtain a critical signal;

[0056] The specific working process is:

[0057] First, the first prediction coefficient, the second prediction coefficient and the third prediction coefficient are marked as yc1, yc2 and yc3 respectively, and then the processing period T4 is set, and a four-dimensional comprehensive prediction vector <first prediction coefficient yc1, second prediction coefficient yc2, third prediction coefficient yc3, processing period T4> is constructed, and then the first prediction coefficient yc1 in the processing period T4 is obtained, and then the average value and fluctuation value corresponding to the first prediction coefficient yc1, the second prediction coefficient yc2 and the third prediction coefficient yc3 in the processing period T4 are statistically analyzed. For any time period Ta in the processing period T4, the average values ​​of the first prediction coefficient yc1, the second prediction coefficient yc2 and the third prediction coefficient yc3 marked in the time period Ta are Pj respectively. yc1 , Pj yc2 and Pj yc3 , and the fluctuation values ​​are Bd yc1 , Bd yc2 and Bd yc3 , and construct 8 different judgment vectors according to different combinations, namely:

[0058] <Pj yc1 +Bd yc1 , Pj yc2 +Bd yc2 , Pj yc3 +Bd yc3 >;

[0059] <Pj yc1 +Bdyc1 , Pj yc2 +Bd yc2 , Pj yc3 -Bd yc3 >;

[0060] <Pj yc1 +Bd yc1 , Pj yc2 -Bd yc2 , Pj yc3 +Bd yc3 >;

[0061] <Pj yc1 -Bd yc1 , Pj yc2 +Bd yc2 , Pj yc3 +Bd yc3 >;

[0062] <Pj yc1 -Bd yc1 , Pj yc2 +Bd yc2 , Pj yc3 -Bd yc3 >;

[0063] <Pj yc1 +Bd yc1 , Pj yc2 -Bd yc2 , Pj yc3 -Bd yc3 >;

[0064] <Pj yc1 -Bd yc1 , Pj yc2 -Bd yc2 , Pj yc3 +Bd yc3 >;

[0065] <Pj yc1 -Bd yc1 , Pj yc2 -Bd yc2 , Pj yc3 -Bd yc3 >;

[0066] Then, any judgment vector is substituted into the fourth prediction model. When the result corresponding to any judgment vector is less than the threshold stored internally in the server, an emergency signal is output and displayed. Upon receiving the emergency signal, the server immediately edits the text and displays "The critically ill patient is in critical condition". Medical staff immediately provide emergency medical treatment to the critically ill patient according to the text content.

[0067] In summary of the above technical solutions: the present invention first collects the clinical data of critically ill patients to be predicted through an information acquisition unit, and then performs predictive analysis on the clinical data of critically ill patients to be predicted through a server based on a pre-established machine learning model to obtain a first prediction result, and at the same time obtains a second prediction result of predictive analysis on the clinical data of critically ill patients to be predicted in a manual participation manner, and then trains the machine learning model based on the first prediction result and the second prediction result, thereby achieving an improvement in the machine learning model, improving the efficiency of data prediction to a certain extent, and ensuring accurate prediction in redundant and complex situations; and then performing hierarchical prediction on the clinical data of critically ill patients to be predicted through at least three levels of pre-trained prediction models, further refining the accuracy of the prediction model, and narrowing the range of the degree, which is conducive to medical staff making accurate judgments on the criticality of critically ill patients.

[0068] The size of the interval and threshold is set for the convenience of comparison. The size of the threshold depends on the amount of sample data and the number of bases set by technicians in this field for each group of sample data; as long as it does not affect the proportional relationship between the parameter and the quantized value;

[0069] The above formulas are all dimensionless and numerical calculations. The formula is a formula obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0070] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0071] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A clinical data analysis and prediction system for critically ill patients based on machine learning, characterized by: It includes a server, an information acquisition unit and an analysis and prediction unit; The information acquisition unit sets a sampling period, and acquires the clinical data of the critically ill patients to be predicted in the server according to the preset sampling period; wherein the clinical data of the critically ill patients include the symptoms of the critically ill patients before and after diagnosis and treatment and the disease diagnosis conclusion; The disease prediction unit performs predictive analysis on the clinical data of the critically ill patients to be predicted based on the pre-established machine learning model through the server to obtain a first prediction result, and at the same time obtains a second prediction result of the predictive analysis on the clinical data of the critically ill patients to be predicted in a manual participation manner; and then trains the machine learning model according to the first prediction result and the second prediction result to obtain an updated prediction model for the next prediction; wherein the first prediction result is a first data set that meets the preset standard and a second data set that does not meet the preset standard obtained by the server based on the pre-established machine learning model to predict the clinical data of the critically ill patients to be predicted; the second prediction result is a third data set that meets the preset standard and a fourth data set that does not meet the preset standard obtained by predicting the clinical data of the critically ill patients to be predicted in a manual participation manner; the disease prediction unit uses at least three levels of pre-trained prediction models to perform hierarchical prediction on the clinical data of the critically ill patients to be predicted to obtain prediction results corresponding to the patient's disease; and also includes a first prediction module, a second prediction module and a third prediction module; The first prediction module uses a pre-trained prediction model to perform a first-level prediction on the clinical data of the critically ill patient to be predicted, and obtains the corresponding first-level prediction coefficient; The second prediction module uses a pre-trained prediction model to perform a second-level prediction on the clinical data of the critically ill patients to be predicted based on the first-level prediction coefficient, and obtains the corresponding second-level prediction coefficient; The third prediction module uses a pre-trained prediction model to perform a third-level prediction on the clinical data of the critically ill patient to be predicted based on the second prediction coefficient, and obtains a corresponding third-level prediction coefficient; The specific process of training the machine learning model according to the first prediction result and the second prediction result is: First, perform data misjudgment analysis based on the first prediction result and the second prediction result, assign corresponding hash values ​​based on the misjudgment analysis results, and match unique data identifiers for corresponding data, specifically: misjudge data that meets the preset standard based on the pre-established machine learning model as data that does not meet the preset standard, and generate a first misjudgment identifier; misjudge data that does not meet the preset standard based on the pre-established machine learning model as data that meets the preset standard, and generate a second misjudgment identifier; misjudge data that cannot be identified based on the pre-established machine learning model but actually meets the preset standard, and generate a third misjudgment identifier; misjudge data that cannot be identified based on the pre-established machine learning model but actually does not meet the preset standard, and generate a fourth misjudgment identifier; wherein, data that cannot be identified based on the pre-established machine learning model but actually does not meet the preset standard is marked as a fifth data set; Acquire data that does not exist in the first data set in the third data set to obtain a sixth data set; and simultaneously acquire data that does not exist in the third data set in the first data set to obtain a seventh data set; and then use the sixth data set and the first identifier, and the seventh data set and the second identifier as input information to train a machine learning model to obtain an updated first prediction model; Acquire data that does not exist in the fourth data set in the second data set to obtain an eighth data set; and simultaneously acquire data that does not exist in the second data set in the fourth data set to obtain a ninth data set; and then use the eighth data set and the first identifier, and the ninth data set and the second identifier as data information to train a machine learning model to obtain an updated second prediction model; Acquire data that does not exist in the first data set and the fifth data set in the third data set to obtain a tenth data set; acquire data that does not exist in the third data set and the fifth data set in the first data set to obtain an eleventh data set; acquire data that exists in the third data set in the fifth data set to obtain a twelfth data set; acquire data that exists in the fourth data set in the fifth data set to obtain a thirteenth data set; and then use the tenth data set and the first identifier, the eleventh data set and the second identifier, the twelfth data set and the third identifier, and the thirteenth data set and the fourth identifier as input information to train a machine learning model to obtain an updated third prediction model; Acquire data that does not exist in the fourth data set and the fifth data set in the second data set to obtain a fourteenth data set; acquire data that does not exist in the second data set and the fifth data set in the fourth data set to obtain a fifteenth data set; acquire data that does not exist in the second data set and the fifth data set in the fourth data set to obtain a fifteenth data set; acquire data that does not exist in the second data set and the fifth data set in the fourth data set to obtain a fifteenth data set; acquire data that does not exist in the second data set and the fifth data set in the fourth data set; and acquire data that does not exist in the second data set and the fifth data set. The pre-trained first prediction model is used to perform the first-level prediction of the clinical data of critically ill patients. The specific working process is as follows: A processing cycle T1 is set, clinical data of critically ill patients to be predicted in the processing cycle T1 is marked as first target data, the first target data is input into a first prediction model, a first prediction result is obtained, and the first prediction result is converted and analyzed to generate a first prediction coefficient; wherein the first prediction result describes the degree of the diagnosis result of the critically ill patient, and the first prediction coefficient is a degree value obtained according to the degree of the disease corresponding to the first prediction result; The pre-trained second prediction model is used to perform the second-level prediction on the clinical data of critically ill patients. The specific working process is as follows: Based on the first prediction result, a processing cycle T2 is set, and the clinical data of the critically ill patient to be predicted in the processing cycle T2 is marked as the second target data, the second target data is input into the second prediction model, a second prediction result is obtained, and the second prediction result is converted and analyzed to generate a second prediction coefficient; wherein the second prediction result is a result of subdividing the degree of the disease based on the first prediction result, the second prediction result describes the degree of the diagnosis result of the critically ill patient, and the second prediction coefficient is a degree value obtained according to the degree of the disease corresponding to the second prediction result; The pre-trained third prediction model is used to perform the third-level prediction on the clinical data of critically ill patients. The specific working process is as follows: Based on the second prediction result, a processing cycle T3 is set, and the clinical data of the critically ill patient to be predicted in the processing cycle T3 is marked as the third target data, the third target data is input into the third prediction model, a third prediction result is obtained, and the third prediction result is converted and analyzed to generate a third prediction coefficient; wherein the third prediction result is the result of subdividing the degree of the disease based on the second prediction result, the third prediction result describes the degree of the diagnosis result of the critically ill patient, and the third prediction coefficient is the degree value obtained according to the degree of the disease corresponding to the third prediction result; Based on the first-level prediction coefficient, the second-level prediction coefficient and the third-level prediction coefficient, the pre-trained fourth prediction model is used to perform comprehensive prediction analysis on the clinical data of critically ill patients. The specific working process is as follows: First, the first prediction coefficient, the second prediction coefficient and the third prediction coefficient are marked as yc1, yc2 and yc3 respectively, and then the processing period T4 is set, and a four-dimensional comprehensive prediction vector <first prediction coefficient yc1, second prediction coefficient yc2, third prediction coefficient yc3, processing period T4> is constructed, and then the first prediction coefficient yc1 in the processing period T4 is obtained, and then the average value and fluctuation value corresponding to the first prediction coefficient yc1, the second prediction coefficient yc2 and the third prediction coefficient yc3 in the processing period T4 are statistically analyzed. For any time period Ta in the processing period T4, the average values ​​of the first prediction coefficient yc1, the second prediction coefficient yc2 and the third prediction coefficient yc3 marked in the time period Ta are Pj respectively. yc1 , Pj yc2 and Pj yc3 , and the fluctuation values ​​are Bd yc1 , Bd yc2 and Bd yc3 , and construct judgment vectors according to different combinations; The judgment vector is then substituted into the fourth prediction model. When the result corresponding to any judgment vector is smaller than the threshold value stored internally in the server, an emergency signal is output to determine that the critically ill patient is at a critical level.

Images